Offline and back into the dark

Although FOSTA, or the Allow States and Victims to Fight Online Sex Trafficking Act of 2017, was just signed on April 11th, the impact of the law has been immediate — and dangerous.

Heralded by supporters as a strong stance against exploitation and sex trafficking, FOSTA dismantled Section 230 of the Communications Decency Act, making it possible for websites and digital platforms to be held legally accountable for sex work-related content posted to their services. Even before it was signed into law, websites were being taken down and terms of service were changed. Shortly after FOSTA became official, notorious escort advertising site Backpage.com was seized by authorities and shuttered.

But FOSTA has also put sex workers in danger, and many have faced serious real-world consequences in the wake of this digital upheaval. Although no official reports have been released as of this writing, anecdotal evidence is trickling in. Johanna Breyer, interim executive director and co-founder of the Saint James Infirmary, a health clinic that supports sex workers in California’s Bay Area, told me that in the weekend following FOSTA, the infirmary’s mobile van outreach saw a dramatic increase of street-based sex workers in the Mission District. Breyer estimated that there were about double or triple the usual number of workers seeking assistance.

Fancy, a Midwestern sex worker who manages a fund dedicated to providing financial support for sex workers in need, has seen a dramatic uptick in requests for help. In the wake of the Backpage shutdown, she says she went from receiving occasional requests for help to a dozen or two dozen messages daily. Many messages were from sex workers asking for advice on how to work on the streets safely. “I never used to have people asking me how to stay safe on the street, or even where to advertise or how to screen,” Fancy tells me. Now, responding those types of requests is a regular part of her routine.

How did FOSTA — a bill pitched as a way to end trafficking and make women safer — wind up driving women to work on the street, putting their lives and safety at risk in the process? The answer lies in the untold story of online sex work, a tale of how modern technology transformed the world’s oldest profession, offering sex workers increased agency, autonomy, and substantially improving their safety.

There are three things that can dramatically improve the safety of sex work, even in places where it’s partially or fully criminalized. As Shannon Kowalski, director of advocacy and policy for the International Women’s Health Coalition, tells me, those key safety strategies are being able to work indoors, being able to screen potential clients, and having access to a community of other sex workers who share information and resources with one another (including databases of dangerous or disrespectful clients commonly referred to as “bad date lists”).

Before sex work moved online, if a person wanted to avoid walking the streets, they needed access to essential safety resources that were largely mediated by third parties like agencies and pimps. It was these third parties that would find and screen clients and deter bad client behavior by making their presence known. If a client proved to be dangerous or untrustworthy, their name would wind up on a blacklist — but in the era before Google Drive, those blacklists were literal spreadsheets, mostly shared locally between agencies (if they were shared at all). If you were fortunate enough to work for someone caring and respectful, this setup could be a wonderful one. But for others, the choice could be working with an exploitative, abusive agency or pimp, or not working at all.

“You really couldn’t make it as an independent unless you were a stripper or did street work,” says Mel, an escort who got her start in New England in the mid-1990s. Some workers who started with an agency did eventually go independent by poaching their favorite regulars and relying on referrals. (That’s how Mel went independent initially.) But it was a strategy that left workers heavily reliant on a handful of hopefully loyal clients.

The internet opened up a new world of possibilities. Advertising boards like the now-defunct Craigslist Erotic Services section enabled sex workers to connect directly with potential clients from the safety of their homes, rather than soliciting business on a street corner or at a strip club, or relying on a third party to connect them with work.

As more sex workers got online, the safety resources continued to improve. In her early days as an independent, Mel would screen clients by asking for their full name and phone numbers. She would cross-reference that with the phone book listing, relying on a potential client’s willingness to be honest — and the implicit suggestion that access to their contact details might offer her some recourse if they acted badly — as a way to weed out more dangerous clients. Maggie McNeill, who ran an escort agency in New Orleans in the early aughts, described a similar process of verification via phone number, noting that a client could be accepted or rejected based on how cop-friendly the hotel they were calling from was known to be.

Online, the phone book could be swapped for global records databases like Intelius and Pipl, with sites like Facebook and LinkedIn providing additional ways to confirm that a client was who they claimed to be. Sex work-specific sites cropped up as well, combining identity verification tools with crowdsourced bad date lists. That gave sex workers around the country (and, potentially, around the world) a way to warn each other about dangerous clients, and it prevented those clients from escaping a bad reputation by moving to a different city.

In the same way that it’s helped bring together all sorts of isolated people with niche interests, the internet proved to be a particularly valuable resource for building a sex work community. Facebook, Twitter, Reddit, and other online forums have all offered ways for sex workers to connect with one another, sharing professional knowledge, survival tips, and even just giving one another essential emotional support.

“Working in isolation is one of the most dangerous [ways] to work,” Maxine Holloway tells me, going on to say that she “couldn’t imagine doing this type of work without community.” While screening resources and bad date lists are obviously essential aspects of sex worker safety, being able to connect with people who understand the various stresses that come with sex work can be equally life-saving. “When you have a stigmatized profession, it is really important to have people who understand what that feels like,” says Holloway.

As sex work moved online, the public perception of the industry began to shift, becoming less taboo in certain corners of society. That sense of increased safety and acceptability may have helped fuel the rise of a more culturally palatable version of sex work marked by high-end escorts who were largely white, privileged, college-educated, and who saw clients, not out of financial need, but out of a love of sex and money and luxury. Such characters popped up on TV shows like The Secret Diary of a Call Girl and The Girlfriend Experience.

At the same time, the shift toward internet-enabled sex work was also a major boon to low-income sex workers. Low-priced or even free advertising sites like Craigslist and Backpage made it easy and affordable to advertise services online, and they provided a vastly safer alternative to street-based work. As the impact of FOSTA continues to ripple out across the industry, it’s those same survival sex workers that will suffer most, as the advertising sites and other safety resources that continue to exist will begin raising their price, effectively shutting out the most vulnerable members of the sex work community.

Sarah, who turned to survival sex work during a homeless stint in spring 2016, is adamant that the internet offered her a lifeline, keeping her afloat during an extremely uncertain time. Sarah was living out of her car with an abusive, heroin-addicted partner, and she needed regular influxes of cash to help her partner manage his drug habit. Sex work was an obvious way to quickly get access to that cash.

Her initial plan involved knocking on truck doors at the gas station where she was currently camped out. It was a strategy that, she now recognizes, could have easily gotten her assaulted or even killed. Fortunately, one of her friends told her about Backpage, and for the next few months, Sarah used the site to connect with clients. Stuck in a precarious and dangerous circumstance, the platform offered at least a modicum of safety, control, and security.

Connecting with clients online meant she could chat with them first and weed out the worst ones, rejecting those she didn’t like without fear for her physical safety. Posting advertisements online could sometimes result in multiple offers of work, giving her the ability to choose the best job rather than take whatever opportunity presented was presented to her.

Even street-based sex workers who aren’t personally using the internet have directly benefited from the internet, specifically from online organizing and activism done on their behalf. “I’m working on an outreach program for sex workers right now,” Holloway tells me. “We are using the internet to find out where outdoor workers are and what their needs are.” Yet, under FOSTA’s expansive definition of “trafficking,” even work like this could be at risk.

The intention of FOSTA may have been to end sex trafficking, or at least drive it off the internet, but no one I spoke with actually thought that would be the end result — nor did anyone think this would be the end of sex work. “People need to make money in some way,” says Kristina Dolgin, a sex worker and founder of Red Light Legal, an organization that provides direct legal services, legal representation, community education, and effective policy advocacy to sex workers in all corners of the industry. When sex work feels like the best option, it’s the one that people will turn to, regardless of whether they’re able to advertise online or if they have access to screening resources.

What’s less clear, however, is what the sex work of the future is going to look like. Some of the people I spoke with predicted a return to pre-internet strategies, where third parties provide access to clients and protection, noting that agencies and pimps had already been coming out of the woodwork to offer their services to independent sex workers in the wake of the Backpage shutdown.

Others were less convinced that a shift back to older models of sex work safety was likely or even possible. “Massage houses and agencies still use these websites to advertise,” says Arabelle Raphael, noting that the now-defunct RedBook was extremely popular with massage parlors. “Everybody uses [online sex work resources], they’re not just for independent sex workers.”

McNeill is similarly bearish about a return to the agency model, telling me that the same anti-trafficking activism that led to FOSTA has resulted in increased crackdowns on agency owners. She points to Amber Batts, a woman she describes as “a really good, ethical, helpful agency owner,” as an example of how things could go wrong for agency owners. Whatever rave reviews Batts may have gotten from the women she worked with didn’t matter to the court that convicted her of sex trafficking. Who would want to take on the work and headache of establishing and running an agency if it could easily mean landing in prison?

What seems most likely to happen is an evolution toward a new model, one that combines established, mainstream search tools with enhanced cybersecurity and sex work-specific sites hosted on offshore servers taking the knowledge gained from the past 20 years of internet-enabled sex work and adapting it to fit the new legal environment.

But getting to that point will take time — time that the most vulnerable sex workers don’t have. And as sex workers struggle to adapt to the realities of a significantly less open internet, it’s the very population that FOSTA purported to protect that’s most likely to lose access to the resources that were keeping them safe.

In Egypt, dating apps are a refuge for a persecuted LGBTQ community, but they can also be traps

Firas knew something was wrong when he saw the checkpoint. He was meeting a man in Dokki’s Mesaha Square, a tree-lined park just across the Nile from Cairo, for what was supposed to be a romantic rendezvous. They had met online, part of a growing community of gay Egyptians using services like Grindr, Hornet, and Growler, but this was their first time meeting in person. The man had been aggressive, explicitly asking Firas to bring condoms for the night ahead. When the day came to meet, he was late — so late that Firas almost called the whole thing off. At the last minute, his date pulled up in a car and offered to take Firas directly to his apartment.

A few blocks into the ride, Firas saw the checkpoint, a rare occurrence in a quiet, residential area like Mesaha. When the car stopped, the officer working the checkpoint talked to Firas’ date with deference, almost as if he were a fellow cop. Firas opened the door and ran.

“Seven or eight people chased me,” he later told the Egyptian Initiative for Personal Rights, a local LGBT rights group. “They caught me and beat me up, insulting me with the worst words possible. They tied my left hand and tried to tie my right. I resisted. At that moment, I saw a person coming from a police microbus with a baton. I was scared to be hit on my face so I gave in.”

He was taken to the Mogamma, an immense government building on Tahrir Square that houses Egypt’s General Directorate for Protecting Public Morality. The police made him unlock his phone so they could check it for evidence. The condoms he had brought were entered as evidence. Investigators told him to say he had been molested as a child, that the incident was responsible for his deviant sexual habits. Believing he would be given better treatment, he agreed — but things only got worse from there.

He would spend the next 11 weeks in detention, mostly at the Doqi police station. Police there had printouts of his chat history that were taken from his phone after the arrest. They beat him regularly and made sure the other inmates knew what he was in for. He was taken to the Forensic Authority, where doctors examined his anus for signs of sexual activity, but there was still no real evidence of a crime. After three weeks, he was convicted of crimes related to debauchery and sentenced to a year in prison. But Firas’ lawyer was able to appeal the conviction, overturning it six weeks later. Police kept him locked up for two weeks after that, refusing to allow visitors and even denying that he was in custody. Eventually, the authorities offered him an informal deportation — a chance to leave the country, in exchange for signing away his asylum rights and paying for the ticket himself. He jumped at the chance, leaving Egypt behind forever.

It’s an alarming story, but a common one. As LGBTQ Egyptians flock to apps like Grindr, Hornet, and Growlr, they face an unprecedented threat from police and blackmailers who use the same apps to find targets. The apps themselves have become both evidence of a crime and a means of resistance. How an app is built can make a crucial difference in those cases. But with developers thousands of miles away, it can be hard to know what to change. It’s a new moral challenge for developers, one that’s producing new collaborations with nonprofit groups, circumvention tools, and a new way to think about an app’s responsibility to its users.

Most arrests start the same way as Firas’ story. Targets meet a friendly stranger on a gay dating site, sometimes talking for weeks before meeting in person, only to find out they’re being targeted for a debauchery case. The most recent wave of arrests started last September after an audience member unfurled a gay pride flag at a rock concert, something the regime took as a personal insult. More than 75 people were arrested on debauchery charges in the weeks that followed.

Homosexuality isn’t illegal in Egypt, but the LGBTQ community has become a useful scapegoat for the el-Sisi regime, and the General Directorate for Protecting Public Morality is being used to jail and prosecute anyone perceived as committing a transgression. Even when the charges don’t stick, charges can be used as a pretense for public humiliation, weeks of imprisonment, or even deportation. The Egyptian Initiative for Personal Rights (EIPR) has documented more than 230 LGBTQ-related arrests from October 2013 to March 2017, which is more than in the previous 13 years combined.

For those in the community, the threat of violence is hard to escape. “I froze as a human being for a while,” one Egyptian called Omar told me. “I lost my sexual drive for a long time. There were so many horrific stories about people being imprisoned or blackmailed or put under some sort of pressure for their sexuality. It was disturbing.”

Egypt’s state media has largely cheered on the crackdown, treating a 2014 raid on the Bab al-Bahr bathhouse as more of a tabloid drama than a human rights issue. Raids on bars, house parties, and other gay spaces have become common. “There’s this sense of society wanting to publicize anything that’s private for the LGBTQ community,” Omar says. “It becomes hard to discriminate what’s private and what’s public.”

As a result, channels for private communications like dating apps Grindr and Hornet are particularly important here. And to different extents, both platforms feel that they have some responsibility for keeping their users safe. In the weeks after the September crackdown, both Grindr and Hornet began sending out warnings through their apps, notifying users of the crackdown and giving the same advice about retaining a lawyer and watching for police accounts. The messages served as a kind of early warning system, a way to spread news of the new threat as quickly as possible.

Since 2014, Grindr has warned Egyptian users about blackmailers and recommended keeping their account as anonymous as possible. If you check the app in Cairo, you’ll see a string of anonymous pictures. Some users even create profiles to warn others that a specific individual is a blackmailer or a cop. On Hornet, more than half the accounts have pictures, though many stay obscured. One Egyptian man told me that when he visited Berlin on vacation, he was shocked to see that every Grindr profile had a face; it had never occurred to him that so many people might out themselves online.

Local LGBTQ groups have their own recommendations for staying safe. Before meeting up, they suggest you have a designated attorney from one of the local groups, and that you tell someone where you’re going in case you get picked up by police. Don’t keep screenshots on your phone or on cloud services like Google Photos that might be accessible to police. If you use video chat instead of sending pictures, it’s harder to take incriminating screenshots. Screenshots are dangerous for the people who take them, too: a Grindr shot in your camera roll could easily become evidence in a debauchery case. Just having the app on your phone is a risk.

It’s good advice, but it’s hard to follow. Even if you know all the rules, all it takes is one slip to fall into the trap. A local nonprofit worker named Youssef told me he tells friends not to use the apps if they have other options. By now, he’s used to being ignored. “It’s mental torture,” he said. “It’s a daily struggle because you just want to express your sexuality.”

It’s easier if the safeguards are built into the app itself. Grindr still collects user locations in Egypt and ranks nearby users from closest to farthest, but the Egyptian version of the app won’t list precise distances. At the same time, Grindr has struggled with a string of recent security issues, leaking profile data through third-party plugins and sharing HIV statuses with analytics partners. None of those slip-ups seem to have been exploited by Egyptian groups, but they can hardly be reassuring to users.

Hornet, Grindr’s main competitor in Egypt, makes no effort to hide a user’s location in Egypt at all. Hornet president Sean Howell told me it was a deliberate choice. “Can someone go through and look for men nearby in Egypt? Yes, they can,” Howell said. “We talk about it. We send warnings. But we have 100,000 users in Cairo. They’re not going to arrest all these men. Are we going to send them back to a digital closet?”

One of the biggest challenges in designing these features is the culture gap between users like Firas and the designers at Grindr and Hornet. Grindr was founded by an Israeli immigrant who settled in LA; Hornet splits its executive team between San Francisco, Toronto, and New York. Both apps were built amid a thriving, sex-positive gay culture. In most countries, they represent that culture pushed to its limit. For Americans, it’s hard to imagine being afraid to show your face on such an app. It’s not just a technological challenge, but a cultural one: how do you design software knowing that simple interface decisions like watermarking a screenshot could result in someone being arrested or deported? Thousands of miles away from the most vulnerable users, how would you know if you made the wrong choice?

Researchers who are partnering with platforms have been struggling with those questions for years, and apps like Grindr have given researchers a new way to answer them. In places where the gay community has been driven underground, dating apps are often the only way to reach them — something that’s led a number of nonprofits to seek out Grindr as a research tool.

“So many guys will get on Grindr who have never told anyone they’re gay,” says Jack Harrison-Quintana, the director of Grindr’s social-good division, Grindr For Equality. “And they know nothing. There’s no network. Once we start messaging them, it creates more of a network.” Harrison-Quintana’s first major project saw Grindr pushing out messages to Syrian refugee arrival areas in Europe, telling new arrivals about LGBTQ resources in the area. Once he saw how powerful the geo-targeted messages could be, he started looking for more places to use them.

In 2016, a human rights NGO called Article 19 came to Harrison-Quintana with a proposal: a massive survey of Grindr’s most vulnerable users, funded by grants and sent out through Grindr’s direct messaging system and supplemented with local surveys and focus groups. The project would focus on three Middle Eastern countries with different degrees of repression: Egypt, Iran, and Lebanon. Egypt faced the most intense crackdown, but the threat had more to do with police intimidation than actual convictions. Iran faces a more subtle version of the same threat, with police more interested in cultivating informants than raiding bathhouses and making headlines. Lebanon is seen as one of the best places to be gay in the region, even though homosexuality is still illegal there. The greatest threat is being accidentally outed at a military checkpoint and swept up in a broader counterterrorism effort.

The project culminated in an 18-person roundtable the following summer, bringing together representatives from Grindr, Article 19, local groups like EIPR, and digital rights technology groups like Witness and the Guardian Project. After Article 19 and local groups presented the results of the survey, the group puzzled through a series of possible fixes, voting on them one by one.

“It was a very democratic meeting,” said Article 19’s Afsaneh Rigot. “I was talking about things we’d seen groups find useful in the past. The local groups were talking about what they think could help their community. The technologists were talking about the features that they could help create. And then people like Jack [Harrison-Quintana] from the business side were talking about what companies would be able to take on.”

The end result was a list of recommendations, some of which are already showing up in Grindr. Since October, Grindr users in 130 countries have been able to change the way the app appears on the home screen, replacing the Grindr icon and name with an inconspicuous calculator app or other utility. Grindr also now features an option for a PIN, too, so that even if the phone is unlocked, the app won’t open without an additional passcode. If you’re stopped at a checkpoint (a common occurrence in countries like Lebanon), police won’t be able to spot Grindr by flipping through your phone. And if co-workers or suspicious parents do catch on to the masked app, they won’t be able to open it without your permission. It’s a small change — one many users in Egypt haven’t even noticed — but it’s a serious step forward for Article 19’s broader project.

Other recommendations were harder to implement. The group suggested that apps would be safer with disappearing messages or images that were harder to screenshot, but making that change might cut too deep into the service itself. It would be easier to slip a debauchery case if those screenshots went to an in-app gallery instead of the phone’s camera roll, but doing so would confuse a lot of users and require deep changes in how the app is engineered. The biggest ask was a panic button, which would let users erase the app and contact friends with a single button press if they realize they’ve been entrapped. So far, no app has built in that kind of feature, and it’s not hard to see why. For every real user in danger, there would be 10 accidental account wipes. It would make users safer, but would it be worth the friction? In the background, there is an even harder question: why is it so hard for tech companies to take stock of this kind of risk?

For Dia Kayyali, a Witness program manager, the problem is built into the apps themselves — developed in cultures without the threat of being jailed or tortured for one’s sexual orientation. “It’s much more difficult to create an app that functions well for gay men in the Middle East,” Kayyali told me. “You have to address the fact that governments have people who are specifically manipulating the platform to hurt people, and that’s a lot more work.” With founders focused on growing first and asking questions later, they often don’t realize what they’re taking on until it’s too late.

“What I would like is for platforms to be designed for the most marginalized users, the ones most likely to be in danger, the ones most likely to need strong security features,” Kayyali said. “But instead, we have tools and platforms that are built for the biggest use cases, because that’s how capitalism works.”

Pulling out of countries like Egypt would certainly make business sense: none of the countries involved are lucrative ad markets, particularly when you factor in the cost of developing extra features. But both apps are fully convinced of the value of the service they’re providing, even knowing the dangers. “In countries where it’s unsafe to be gay, where there are no gay bars, no inclusive sports teams, and no queer performance spaces, the Grindr app provides our users with an opportunity to find their communities,” Quintana-Harrison told me. Leaving would mean giving that up.

When Howell visited Egypt in December for Hornet, he came away with a similar conclusion. Hornet has made some small security changes since the trip, making it easier to add passwords or delete pictures, but the bulk of his work was telling users what was happening and pressuring world leaders to condemn it. “[Egyptian users] don’t want us to shut down,” he told me. “Gay men will not go back into the closet. They’re not going to abandon their lives. They’re not going to abandon their identity even in the harshest conditions. That’s what you’re seeing in Egypt.”

He was more skeptical about the value of the new security measures. “I think a false sense of security can put users in harm’s way,” Howell said. “I think it’s far more important to teach them about what the situation really is and make sure they’re aware of it.”

That leaves LGBTQ Egyptians with a fear that can build up in unexpected ways. It hit Omar a few weeks after the first raids this fall. It felt like there was a new arrest every day, and no place left that was safe. “I was walking down the street, and I felt like there was someone following me,” he told me. When he turned around to check, there was no one there. “It was in that moment that I realized I am afraid for my life. The situation is not safe here in Egypt. It’s actually dangerous. And then I decided, if it’s actually dangerous, then it’s time to speak out.”

It was supposed to be the laptop that saved the world.

In late 2005, tech visionary and MIT Media Lab founder Nicholas Negroponte pulled the cloth cover off a small green computer with a bright yellow crank. The device was the first working prototype for Negroponte’s new nonprofit One Laptop Per Child, dubbed “the green machine” or simply “the $100 laptop.” And it was like nothing that Negroponte’s audience — at either his panel at a UN-sponsored tech summit in Tunis, or around the globe — had ever seen.

After UN Secretary-General Kofi Annan offered a glowing introduction, Negroponte explained exactly why. The $100 laptop would have all the features of an ordinary computer but require so little electricity that a child could power it with a hand crank. It would be rugged enough for children to use anywhere, instead of being limited to schools. Mesh networking would let one laptop extend a single internet connection to many others. A Linux-based operating system would give kids total access to the computer — OLPC had reportedly turned down an offer of free Mac OS X licenses from Steve Jobs. And as its name suggested, the laptop would cost only $100, at a time when its competitors cost $1,000 or more.

“We really believe we can make literally hundreds of millions of these machines available to children around the world,” Negroponte promised. “And it’s not just $100. It’s going to go lower.” He hinted that big manufacturing and purchasing partners were on the horizon, and demonstrated the laptop’s versatile hardware, which could be folded into a chunky e-reader, a simple gaming console, or a tiny television.

Then, Negroponte and Annan rose for a photo-op with two OLPC laptops, and reporters urged them to demonstrate the machines’ distinctive cranks. Annan’s crank handle fell off almost immediately. As he quietly reattached it, Negroponte managed half a turn before hitting the flat surface of the table. He awkwardly raised the laptop a few inches, trying to make space for a full rotation. “Maybe afterwards…” he trailed off, before sitting back down to field questions from the crowd.

The moment was brief, but it perfectly foreshadowed how critics would see One Laptop Per Child a few years later: as a flashy, clever, and idealistic project that shattered at its first brush with reality.

If you remember the OLPC at all, you probably remember the hand crank. It was OLPC’s most striking technological innovation — and it was pure vaporware. Designers dropped the feature almost immediately after Negroponte’s announcement, because the winding process put stress on the laptop’s body and demanded energy that kids in very poor areas couldn’t spare. Every OLPC computer shipped with a standard power adapter.

By the time OLPC officially launched in 2007, the “green machine” — once a breakout star of the 21st-century educational technology scene — was a symbol of tech industry hubris, a one-size-fits-all American solution to complex global problems. But more than a decade later, the project’s legacy is more complicated than a simple cautionary tale. Its laptops are still rolling off production lines, and a new model is expected later this year.

And people are still talking about the crank.

Nicholas Negroponte was a self-described optimist, and his business was inventing the future. A professor with decades of experience at MIT, Negroponte had co-founded the university’s influential Media Lab in 1985. He’d been one of the first backers of Wired magazine, where he wrote a column evangelizing the transformative power of technology. And he had a longtime passion for education — where computers, he thought, could be revolutionary.

Negroponte believed in constructionism: an educational theory that said children should learn by making things and solving problems, rather than completing worksheets or attending lectures. In 1982, Negroponte and an MIT colleague and key constructionist figure Seymour Papert paired up for an initiative at a French-funded research center in Senegal, teaching children to program on Apple II computers. (Negroponte did not return requests for an interview for this article.)

By the late 1990s, children’s computing initiatives were a major political priority in the States. President Bill Clinton popularized the idea of a “digital divide” between rich and poor, and some American schools began issuing students individual computers to close the gap. Microsoft and Toshiba sponsored a laptop distribution program called Anytime Anywhere Learning, and Maine funded a statewide initiative with input from Papert himself.

Negroponte, however, was more interested in reaching students who might never have seen a laptop at all. In 1999, he and his wife opened a school in the remote Cambodian village of Reaksmei, equipping it with a satellite dish, generators, and rugged Panasonic Toughbook laptops.

It was a key moment for Negroponte. He considered the program a success, talking to reporters about how children would use the laptops as the only source of electrical light in their homes — “talk about a metaphor and a reality simultaneously,” he quipped at one point. But most schools across the developing world couldn’t afford Toughbooks. They needed a new kind of device.

“How do you take these ideas that we’d been — we think — successfully working with for 40 years and bring them to scale?” says Walter Bender, who was Negroponte’s colleague and would go on to be a co-founder of OLPC. “[Nicholas Negroponte] had the insight that the real issue wasn’t a lack of good ideas. The real issue was a lack of access to computers.”

OLPC wasn’t just a laptop, it was a philosophy. Negroponte insisted that kids needed to personally own the computers, so they’d be invested in maintaining them. And they should be able to use them anywhere, not just under teachers’ supervision. For inspiration, he pointed to educational researcher Sugata Mitra’s famous Hole-In-The-Wall experiment, where children taught themselves to use a computer in a Delhi slum. Mitra’s vision was more minimalist than OLPC’s, but both projects were almost totally focused on distributing computers. Kids’ natural curiosity was supposed to do the rest.

This was a provocative idea, and Negroponte’s OLPC pitch toed the line between ambition and hubris. The organization wouldn’t just sell hundreds of millions of units, he declared; it wouldn’t even take orders of less than a million. The laptop wasn’t just tough, it was so tough you could throw it across a room — a feature he’d happily demonstrate in interviews.

These pronouncements made headlines. “We wouldn’t be having this conversation — and OLPC wouldn’t be recognized still, ten years later, by many people — if they hadn’t gone in big from the start,” says Christoph Derndorfer, former editor of entrepreneur Wayan Vota’s now-defunct blog OLPC News.

But OLPC’s overwhelming focus on high-tech hardware worried some skeptics, including participants in the Tunis summit. One attendee said she’d rather have “clean water and real schools” than laptops, and another saw OLPC as an American marketing ploy. “Under the guise of non-profitability, hundreds of millions of these laptops will be flogged off to our governments,” he complained. In the tech world, people were skeptical of the laptop’s design, too. Intel chairman Craig Barrett scathingly dubbed OLPC’s toy-like prototype “the $100 gadget,” and Bill Gates hated the screen in particular. “Geez, get a decent computer where you can actually read the text,” he told reporters.

Even fans of OLPC were somewhat dubious. “We were excited about the prospects, but kind of scared by the over-simplistic plan, or lack of plan,” recalls Derndorfer. OLPC News could have been an enthusiastic booster of OLPC, but it was also a relentless gadfly — its first archived post, quoting a ZDNet report, is titled “OLPC not a good beginning.”

And the laptop Negroponte was pitching in 2005 simply didn’t exist. OLPC’s prototype was little more than a mockup. It hadn’t signed a manufacturer, let alone priced out a sub-$100 product. Groundbreaking technologies like the crank and mesh networking system were still mostly theoretical.

Rabi Karmacharya — whose education nonprofit Open Learning Exchange Nepal runs one of the oldest existing OLPC deployments — says that for people pitching OLPC to their local governments, utopian hype about a $100 self-powering computer wasn’t helpful. It distracted people from the promise of what OLPC was actually building: a tiny, low-power laptop at an incredible price.

Despite the clumsy unveiling in Tunis, within a month OLPC had secured a deal with Taiwanese computer maker Quanta, whose founder Barry Lam liked the project’s humanitarian mission. OLPC announced plans to launch by the end of 2006, shipping a million laptops apiece to seven countries, as well as smaller numbers to developer communities elsewhere. Quanta was even supposed to explore building a commercial version of the laptop.

OLPC had made genuine technical breakthroughs. In its very early concept designs, the laptop used a rear projection screen that gave it a tent-like look; the final product featured a custom LCD display designed by CTO and co-founder Mary Lou Jepsen. The screen toggled between full-color and black-and-white modes, consuming a fraction of the power that a standard display would need. It could be manufactured for just $35, which was more than Negroponte initially wanted, but still remarkably cheap.

OLPC’s first prototype looked like a conventional computer, albeit one that was bright green and book-sized. Lauded designer Yves Béhar soon came on board to rework almost every other aspect of the laptop’s aesthetic. Behar says the team spent almost a year sending prototypes to schools around the world for feedback, as they slowly negotiated a compromise between looks and practicality.

The result was a distinctive-looking machine known as the XO-1: a toylike green-and-white laptop with rounded edges, a swiveling “neck” instead of a standard hinge, and a chunky bezel around its 7.5-inch screen. Every flourish on the XO-1 was designed to serve a purpose. Its screen folded into the keyboard to create a tablet, controlled by a few buttons on its bezel.

Ear-like antennas flipped up to extend its Wi-Fi range, while protecting the laptop’s ports when they were folded down. A decorative XO logo was printed in hundreds of color permutations, so kids could tell their laptops apart. And a dustproof one-piece rubber keyboard made it easy to print any key layout. “Some countries got their very first keyboard ever created in their own local language,” says Béhar.

Walter Bender, meanwhile, was working on a lightweight operating system designed specifically for children. Sugar OS was built on Red Hat Linux, and its open-source design would let kids poke around the laptop’s core firmware. Instead of using the standard desktop computing metaphor, Sugar’s app icons were arranged like a digital charm bracelet — a simple ring of animals, musical notes, and shooting stars. “It was very tool-oriented: tools for doing things, for making things,” recalls Bender. “It wasn’t curriculum-oriented. It wasn’t a bunch of exercises.”

With The New York Times publishing headlines like “The laptop that will save the world,” and millions of sales on the horizon, OLPC looked set for success. Then, everything started to fall apart.

After announcing “the $100 Laptop,” OLPC had one job to do: make a laptop that cost $100. As the team developed the XO-1, they slowly realized that this wasn’t going to happen.

According to Bender, OLPC pushed the laptop’s cost to a low of $130, but only by cutting so many corners that the laptop barely worked. Its price rose to around $180, and even then, the design had major tradeoffs. The XO-1 was easy to take apart — there were even a few spare screws inside its handle. But things like the screen could only be replaced with OLPC-specific parts. The solid-state storage was sturdier than a traditional hard drive, but so expensive that the XO-1 could hold only a single gigabyte of data. Some users complained that the one-piece rubber keyboard fell to pieces after too much typing. The internet-sharing system barely worked, and it was quickly removed from Sugar.

While Sugar was an elegant operating system, some potential buyers were dubious of anything that wasn’t Microsoft Windows. They wanted students to learn an interface they’d be using for the rest of their lives, not just with the XO-1.

OLPC may have undercut even the XO-1’s strong points by overselling them. “The utopianism set unrealistic expectations around what the laptops should be able to accomplish,” says Morgan Ames, a Berkeley researcher who’s currently writing a book about OLPC. That included Negroponte’s laptop-tossing demonstrations. “When you’re talking about a laptop that kids are using surrounded by concrete floors and cobblestone streets — there was a ton of breakage that really blindsided projects, because they expected these laptops to be a lot more indestructible.”

And since OLPC had put so much focus on cost, Bender began to worry that people saw the project as a hardware startup, not an educational initiative. He remembers debating the laptop’s name with Negroponte: instead of “the $100 laptop,” Bender wanted to call it “the Children’s Machine,” he says. “I think we got more mileage out of ‘The $100 Laptop’ at the time, because typical laptops cost over $1,000, so it was a very bold statement. But we got burned by that — because we set an expectation around price, rather than an expectation around what this machine was really for.”

While OLPC was still designing the XO-1, Intel announced that it too was building a cheap educational laptop. The Classmate PC would be small and rugged like OLPC’s design, but run the more familiar Windows XP operating system and sell for somewhere between $200 and $400. As OLPC’s full-scale launch slipped to 2007, and its $100 price tag faded away, Intel shipped the first Classmate PCs to Brazil and Mexico.

Negroponte was furious. He lashed out at Intel, accusing it of approaching OLPC’s target markets and dumping laptops below cost to destroy the nonprofit. “Intel should be ashamed of itself,” he said. “It’s just — it’s just shameless.”

OLPC, which prided itself on not being a tech company, had little experience landing hardware contracts. It announced incredible sales numbers, only to have buyers scale back or drop out. Officials in India, one of OLPC’s original seven customers, reportedly killed the deal because of a long-standing feud with the Media Lab. In a particularly stinging loss, Libya canceled an order of 1.2 million XO-1 laptops and bought Classroom PCs instead.

Bender thinks OLPC might have struck more deals if it had focused less on technical efficiency. “Every conversation we ever had with any head of state — every time — they said, ‘Can we build the laptop in our country?’” he says. “We knew that by making the laptop in Shanghai, we could build the laptop [to be] much less expensive. And what we didn’t realize was that the price wasn’t what they were asking us about. They were asking us about pride, not price. They were asking us about control and ownership of the project.” OLPC had created a computer that could withstand dust and drops, but it hadn’t accounted for political messiness.

As development dragged on, the XO-1 started looking less technically impressive, too. In mid-2007, Taiwanese company AsusTek revealed an eye-catching new computer called the Eee PC, delivering a cheap, tiny laptop without OLPC or Intel’s educational trappings. The Eee PC had a lot of the XO-1’s drawbacks: slow performance, a tiny screen, a minuscule hard drive, and a cramped keyboard. But the $399 machine was an unexpected success. It sold 5 million units in the first year, and other laptop makers quickly released their own “netbook” computers, fueling a massive boom in cheap, tiny laptops.

Meanwhile, OLPC’s own prospects looked increasingly modest. One 2007 report suggested that the XO-1’s first production run would be a paltry 300,000 laptops. The final numbers weren’t quite so bad. OLPC ran a “Give One Get One” program where people paid $400 to buy a laptop for themselves and a student, raising $35 million and selling 162,000 computers. It managed sizable deals with Mexico, Uruguay, and Peru, for a total of around 600,000 XO-1 sales by the end of the year.

Even so, this was far from the original estimate of 5 to 15 million. “I have to some degree underestimated the difference between shaking the hand of a head of state and having a check written,” Negroponte finally admitted. “And yes, it has been a disappointment.”

The launch of the XO-1 should have been just the beginning for OLPC — but for two of the group’s three co-founders, it was nearly the end. At the start of 2008, Mary Lou Jepsen left to found a low-power display company called Pixel Qi. A few months later, OLPC took a step that Walter Bender thought was unforgivable: it compromised its commitment to open source software, partnering with Microsoft to put Windows on the XO-1.

OLPC’s hardware and software sides split up, and Bender kept managing its software under a separate outfit called Sugar Labs. Ultimately, OLPC’s Windows XP model never got beyond a test run, and the laptops use Sugar to this day. “The only thing I did was to stop getting a paycheck,” Bender jokes. “So it wasn’t very smart on my part, but c’est la vie.”

OLPC plunged ahead by announcing a futuristic dual-screened laptop called the XO-2 in May of 2008, but as the US spiraled into the Great Recession, the organization struggled. When it tried to raise money with a second “Give One Get One” sale, it made less than a tenth of its earlier takings. Startled, Negroponte slashed the initiative’s budget, halved its staff, and created two separate organizations to manage it. His own Boston-based “OLPC Foundation” would develop new hardware — though not the XO-2, which was unceremoniously canceled. A Miami-based “OLPC Association,” led by his friend Rodrigo Arboleda, would distribute its existing laptops.

This put OLPC’s base of operations closer to Latin America, where most of its laptops were going. Peru ordered nearly a million XO-1 computers, but its program was plagued with logistical problems, as the machines went to schools with spotty electrical power and teachers got little support or training. A smaller program in Uruguay fared better, distributing 400,000 laptops across the country’s entire primary school population. This was a real win for OLPC, but by the time it happened, many people already considered the initiative a failure.

OLPC’s Microsoft partnership had alienated parts of the open source community and come to nothing. It had experimented with an American program in Birmingham, Alabama, but its key contact — Birmingham’s mayor — was arrested for running a multimillion-dollar bribery ring. In 2009, TechCrunch named OLPC’s laptop one of the biggest product flops of the decade, blaming “corporate infighting and the glare of reality” for its demise.

And Negroponte was losing interest in hardware. After he outlined a dramatic (and ultimately metaphorical) plan to drop tablets out of helicopters, the OLPC Foundation distributed mass-market Motorola Xoom tablets in two Ethiopian villages as a new experiment. In 2012, it reported that children had learned the alphabet within two weeks, and within five months, they had “hacked Android” — which referred to turning off software that disabled the camera. As Android phones and tablets became more sophisticated, Negroponte abandoned development of an OLPC solar-powered XO-3 tablet. He joined the newly founded Global Literacy XPrize soon after, effectively putting OLPC behind him.

Arboleda tried to reboot the OLPC Association and offer more institutional support to schools, but the project’s prospects continued to dim. A 2012 controlled study in Peru found that laptops hadn’t improved children’s math or language skills, although there were some other cognitive skill improvements. Cheap laptops were just one factor in children’s educational opportunities, and with so many different options on the market, OLPC seemed downright obsolete.

OLPC released a low-end American consumer tablet in 2013, slipping a plastic shell over a generic Android device. The same year, it released an updated version of the XO-1 with new components and an optional touchscreen, called the XO-4. But there were no more grand experiments or expansive plans for the future. In early 2014, the Boston-based OLPC Foundation quietly disbanded, and OLPC News shut down. “Let us be honest with ourselves. The great excitement, energy, and enthusiasm that brought us together is gone,” Vota wrote in a farewell post. “OLPC is dead.”

As far as most of the world is concerned, OLPC is dead. But Sameer Verma, an OLPC community leader who sits on Sugar Labs’ oversight board, isn’t discouraged. “Ten years out, I kind of feel like, should I still be doing this?” muses Verma, a professor of information systems at San Francisco State University. “I think it’s still worth it. Because at its very core, this is really about our ability to solve our own problems, right? The more you know, the better off you are.”

Like many remaining OLPC enthusiasts, Verma isn’t dogmatic about hardware. Sugar Labs is a self-contained project at this point, and its apps have been ported to a web-based launcher called Sugarizer, which can run on just about any platform. In addition to laptops, several OLPC programs distribute standalone servers containing articles from Wikipedia, educational videos from the Khan Academy, and customized material from local educational programs. They’re huge repositories of information for kids and teachers without steady internet connections, and can be accessed via any kind of laptop or tablet.

“My mother’s family comes from a small farming village, and so growing up, we were very familiar with the lifestyle there,” says Verma. “When I would look at OLPC, I would be like — I know that this stuff can really do a lot of good things for communities like that.”

He’s a big fan of XO-1 laptops — many of which, despite very real technical problems, are still running after ten years. “It blows my mind that that stuff still works,” he says. “The batteries still work, the Wi-Fi still works, and amazingly, OLPC still cranks out the software images for it.”

In 2015, the OLPC Association was bought by the Zamora Terán Foundation, a nonprofit created by Nicaraguan banking tycoon Roberto Zamora. His son Rodrigo Zamora, an OLPC Association board member, says the new OLPC is focused on getting laptops to non-governmental organizations and updating hardware only when absolutely necessary. It’s currently designing a successor to the XO-4, with a bigger screen and more powerful components — apparently because manufacturers complained that the old parts were getting hard to find. “If it was [up to] us, we would continue with this,” says Zamora. “Basically, our intention is to keep it as similar as possible, because it’s having such good results.”

Some OLPC deployments are still run through governments. Rwanda, for instance, has spent the last decade gradually giving laptops to young students. The project’s coordinator Eric Kimenyi says it’s distributed 275,000 across 1,500 schools, a reach that is expanding as more schools get access to electrical power.

Some projects, like OLE Nepal, work with education departments but are run as nonprofits. Instead of trying to reach an entire country, OLE Nepal has spread around 5,300 laptops across areas where OLPC’s hardware still has an advantage: remote rural districts with no data networks or wired internet, accessible only through hours of hiking.

Other deployments are even smaller and privately run. OLPC volunteer Andreas Gros is currently trying to set up a new project in Ethiopia, providing laptops and servers to a social center for vulnerable children. (Gros says the laptops are currently tied up in customs.) For him, the name is more important than the hardware. “People know what OLPC stood for,” he says. “They may not know the details, but at least it gives them an idea of what you’re trying to do.”

None of these projects follow Negroponte’s original plan. Nepal and Rwanda’s operations don’t provide one laptop for every child, for instance — children use them in shifts, reducing the total cost to schools. The laptops generally stay in the classroom, where they’re easier to protect and maintain. But the initiatives are still working toward OLPC’s larger goal, often using the original XO-1 laptops.

The present-day OLPC Association’s relationship with these projects is tenuous. Zamora says that OLPC personally visits schools to deliver laptops, but Verma and others in the volunteer community say they barely even know what’s happening at the organization. “The gap between the volunteers and the company itself has increased, to a point where there’s hardly any feedback between us and them,” says Verma.

After years of insisting that it wasn’t a tech company, OLPC really has opted out of the laptop arms race, embracing its status as a niche machine. OLPC’s current laptop has the same camera and screen resolution as its original 2008 edition, and less memory and storage than a budget smartphone. OLPC estimates it’s shipped a total of 3 million XO machines over the course of the past decade. “We’re not in the business of selling laptops,” says Zamora. “If we don’t grow 10, 15, 20 percent a year, that doesn’t matter for us.”

So why keep building the XO at all? That’s the question that drew Negroponte away from OLPC, but not one that bothers Zamora. “With a little money, we can have a lot of impact on poor communities throughout the world,” he says. “[Other laptops] need to get replaced after weeks of being in the field, with the dust, the water, the heat.” And though some phones and tablets are cheaper in the short term, one rugged OLPC might outlast generations of them.

To kids who grew up around smartphones and tablets, says Karmacharya, OLPC’s XO design looks hopelessly outdated. “If their parents happen to have even a low-cost smartphone, they’re more interested in that than the laptop.” But the device is tougher than a cheap Android tablet, and its unique design makes it harder to steal. Users can rely on Sugar’s development community to maintain the software. And unlike a phone or tablet, it’s custom-built for making things, not consuming them. “We’re constantly looking out for any sort of alternative,” he says. “And to date, we have not found anything that compares.”

Negroponte has said that OLPC deserves more credit for helping drive down computer prices during the netbook boom. “We estimate that there are about 50 million laptops in the hands of kids who wouldn’t have otherwise gotten them, not because we made those laptops, but because we pulled the prices down,” he said in one interview. “People might remember that little green and white laptop, but the real success was lowered cost worldwide.”

And even today, co-founder Mary Lou Jepsen believes that laptops are vital for education. “Better teacher training only can get you so far when many of the teachers paid to show up don’t and many more are illiterate. Giving children access to information enables them to keep learning, to keep asking ‘why’ and ‘why not.’”

There’s surprisingly little hard data about the long-term impact of OLPCs on childhood education, though. Zamora points to some case studies for individual countries, and says OLPC wants to commission more comprehensive research in the future. But the organization has mostly focused on anecdotes and distribution numbers as markers of success. “OLPC was always very averse to measuring how well they were doing versus the traditional school system,” says Gros. “There have only been a very limited number of attempts to actually measure how well students were doing with OLPC versus not, because it was very hard to do.”

Ames thinks that OLPC’s high-profile failures helped temper the hype around ed-tech programs. “There was a lot of worry that OLPC would crash and take everything with it — that there would be no funding in [educational technology], there would be no funding in tech development,” says Ames. “I think ed-tech in particular can still really draw on some of the same tropes, and hasn’t fully learned the lessons that OLPC should have taught it. But both of those spaces did have to mature to some degree, and stop being quite so naive in their tech utopianism.”

Non-OLPC student laptop programs are still contentious. Maine Governor Paul LePage trashed his state’s initiative as a “massive failure” in 2016, and while it’s still running, its results have been ambiguous and difficult to measure. Mitra’s Hole-in-the-Wall project won a $1 million TED prize in 2013, but critics say he still hasn’t published any rigorous studies of its effects. Bender isn’t convinced that Mitra’s minimalist computing project proved anything. “We already knew that kids could learn to use computers. They’ve been doing that since day one,” he says. “What the project did not demonstrate is that kids could use computers for learning.”

Ames says the real question isn’t whether laptop programs help students, but whether they’re more effective than other programs competing for the same money. “I think that given unlimited funding, absolutely … Learning about technology is very important,” she says. “That said, there’s always a tradeoff. There’s always some project that will be defunded or de-emphasized as a result of this.”

Thirteen years ago, OLPC told the world that every child should get a laptop. It never stopped to prove that they needed one.

Years ago, I was one of the people who bought into One Laptop Per Child’s early hype. I yearned for a cheap computer that I’d never have to plug in. I swooned over its adorable design. (Those little ears!) I vaguely believed the crank was real, even after I saw an XO-1 firsthand without one. I became, and remain, a huge fan of the Eee PC that followed it. But I’d never actually used the laptop until a couple of months ago, when I ordered one off eBay on a whim.

Besides a missing battery, my XO-1 works perfectly, or at least, as perfectly as I could expect from a decade-old computer. I’ve showed off its apps to my colleagues, although it’s so slow that some wander away while they’re launching. Whatever people say about its ruggedness, the hinge feels fragile in my hands. It is eminently a children’s machine, not an all-purpose laptop. My adult brain is already trained on other operating systems, and my fingers barely fit the rubber keys.

But I’ve still never seen anything like it.

A spider brooch-wearing mathematician

On a crisp Saturday morning in Orsay, a southwestern suburb of Paris with some 16,500 inhabitants, the rue de Paris was bustling. But while many residents were doing their usual weekend shopping at the fishmonger or the butcher shop, further up the street, in a small former chateau that is now the town’s cultural center, about 80 people had set aside their late-morning hours to hear the “voeux” of their legislative representative to the National Assembly, Cédric Villani.

The voeux, or “new year’s wishes,” are a standard exercise of French politicians from the president on down, in which they review activities of the past year and lay out projects for the year to come. Villani, a mathematician and Fields Medal winner (often shorthanded as the equivalent of the Nobel Prize in mathematics), was new to the practice; only six months earlier, he was still an academic. He was dressed as always — winter or summer — in a black three-piece suit, a shirt with cufflinks, a spider brooch on his lapel, and a large, floppy tie called a lavallière (today’s version in purple). He cut an unmistakable figure, sporting a three-day beard, his dark hair styled in a pageboy. He mingled, smiling with attendees, and posed for selfies before taking the stage.

The fact that a mathematician could be considered, as he is, a “rock star” — or, better yet, “the Lady Gaga of mathematics” — says perhaps more about the French than Villani. Nonetheless, Villani, 44, has become a darling of President Emmanuel Macron’s young technocratic government, accompanying the president to Ouagadougou, Burkina Faso, in November and Beijing in mid-January. The government has piled the work on his desk, which is evidence, Villani says, of the need for people with scientific expertise in politics. But of all his projects — from math education to the future of New Caledonia to tax evasion — perhaps his most all-consuming mission is his task force on artificial intelligence and the highly anticipated report it’s set to release tomorrow. If successful, the report will help set the AI agenda in France and Europe for years to come.

In view of a world where “artificial intelligence will be everywhere, like electricity,” as Villani has said, becoming a leader in the field is critical for France. Many feel that Europe is already at an enormous disadvantage compared to the US and China and will need to do some Usain Bolt-style sprinting to catch up. For one thing, France and Europe don’t have the data-gathering platforms necessary to fuel machine learning: they lack the power of what the acronym-loving French call GAFA (Google, Apple, Facebook, and Amazon). French bureaucracy has also historically been a drag on entrepreneurship and invention. Compared to the US, cooperation between academia and industry is much less frequent. And though France is known for the quality of its engineers and scientists, much of the top-level talent goes abroad, where there is more money and freedom to pursue research without constraints. Addressing these issues by sketching the nation’s AI road map has fallen on the well-tailored shoulders of Villani.

Several months before delivering his new year’s wishes, after a local TEDx presentation in November on “How AI Will Revolutionize Health,” I sat down with Villani, who was cordial if a bit distant. After winning the Fields Medal, Villani, a self-described “formerly shy” person, took a media training workshop. His large eyes, luminous skin, thin body, and slightly walleyed expression accentuated the impression of speaking to an “extraterrestrial,” as Paris Match once put it.

The report Villani is set to release isn’t a first for France. At the very end of François Hollande’s presidency last spring, his administration released a rushed AI report, offering broad brushstrokes. But, Villani says, his report should offer both “a panorama … and make a diagnosis of the subject…. It must pose questions explicitly and offer practical solutions and implementations.”

Villani’s six-member task force (@MissionVillani) is made up of a machine learning researcher, an engineer with the defense ministry, and four members of a French digital technology advisory council, with expertise in everything from philosophy to law. The group was charged with a broad-ranging mission, covering industrial, data policy, employment and training, environmental, ethical, and research issues. Contrary to the reports issued by the Obama administration (which one international observer notes “have not led to a single bit of U.S. policy”), Villani’s team expects that some concrete measures will be put into play within months. Focusing on four key sectors — health, transportation, environment, and defense — the team also emphasizes that it has devoted substantial attention to the ethics of data policy within the context of the General Data Protection Regulation (GDPR), which is slated to go into effect in May and will broaden privacy protections for individuals in Europe. Considering the current Facebook meltdown over the Cambridge Analytica scandal, the French team has probably chosen wisely. As for the question of financing, while France’s previous AI report suggested a 1.5 billion-euro investment in AI, it is unclear precisely how much funding will be allotted.

Given the breadth and complexity of the subject of artificial intelligence, a skeptic may doubt one man’s ability to understand the field without years of study, but there is probably no other member of Parliament better suited to lead this project on the subject. By all accounts a quick study with an enormous work capacity and a diplomatic, optimistic temperament, Villani was, of course, also a high-level researcher in a related field, which helped him grasp AI concepts quickly.

Asked if he had an interest in artificial intelligence before he was assigned to the task force, Villani enunciated clearly and deliberately in his high-pitched, slightly theatrical French: “The subject had grown so significantly that you would have to have been blind and deaf not to be interested.” In any case, “the big concerns are not really about the most technical issues.” He has indicated in the past that he hopes to avoid AI’s potentially “devastating effects on economic issues and the democratic fabric,” partly by making sure that AI is “everybody’s business.” Hence his large-scale offensive in the French press to educate the public and his push to seek broad-ranging input for his report.

In November, Villani estimated that he would speak to 250 people for the AI report and finish it by the end of January. But, solicited by hundreds of people who wanted their say, he kept speaking to more and more parties. (“He always says yes,” said one of his team members.) Among these were scientists, lawyers, doctors, philosophers, labor union representatives, business leaders, startup entrepreneurs, and even one roundtable of 15 girls, who discussed the involvement of girls in the sciences. In the end, the task force interviewed about 350 people in groups of 10 to 15, gathered according to topic, in an off-white conference room at the Digital Ministry. There were also 1,600 contributors on a public online platform.

Now, the report is slated to be officially delivered to the president at a March 29th ceremony at Paris’ Collège de France, which will be attended by an estimated 500 guests, with the participation of tech-celebrity guests like Facebook’s chief AI scientist and deep-learning guru Yann LeCun.

AI experts in France and abroad are anxious to see the report. “Governments around the world are struggling with whether they have to do something preemptively about AI,” said Joshua Gans, professor of strategic management at the University of Toronto and co-author of the forthcoming Prediction Machines: The Simple Economics of Artificial Intelligence. “There have been a lot of concerns about potential issues, from safety to jobs to what are its privacy implications.”

If he had to sum up international government reaction to AI, Gans said, “it is a great dropping of the ball … not so much on the research side but on ensuring that privacy laws are up to date and moving toward international agreements … on autonomous weapons.” The fact that France has put a “very high-profile person” in charge of the task force attests to how seriously it is taking these issues, he said.

Jean-Gabriel Ganascia, a professor of computer science at the Sorbonne and author of Le mythe de la singularité (“The Myth of the Singularity”), was one of the people interviewed by Villani and his team. “We’re waiting for the report and then its translation into concrete actions,” Ganascia said. “The important thing is not to write reports and create strategies. We have to act.”

And there is not a moment to waste; some maintain that it is already too late for France and Europe. “A country that doesn’t have an AI industry will be underdeveloped tomorrow. It will be a slow process of technical, political and military colonization,” said provocateur Laurent Alexandre, author of La guerre des intelligences (“The Intelligence War”), in an interview with the French business daily Les Echos. Noting the dominance of American and Chinese tech giants, he said, “Europe has completely lost the AI battle.”

Roxanne Varza, the 33-year-old American director of Paris’ new Station F, which bills itself as the world’s biggest startup campus, let out a melodious laugh when I mentioned Alexandre’s grim views. “That is such a passé view,” she said, sitting in a glass-walled room overlooking her monster tech playground. “Maybe you could have said that five years ago, but I don’t think you can say that anymore. With the Brexit climate, Donald Trump, high Silicon Valley prices, the political situation worldwide, and with Macron now in power in France, I think we’ve seen a huge shift.”

In fact, Varza said, the top countries that apply for programs at Station F are the US and the UK, respectively. “France has a huge opportunity in [the AI] space because French engineers and data scientists are so well known.” And feared American tech giants Facebook and Microsoft are right on the Station F campus, apparently eager to tap into some of that je ne sais quoi in Macron’s “Start-up Nation.” (Add to this the announcement made earlier this year about Facebook’s 10 million-euro investment in France for AI research and a Google AI research center in Paris.)

For his part, Villani doesn’t want an AI war; he wants competition, yes, but also collaboration, such as one he set up with Microsoft when he was director of the Institut Henri Poincaré, a prestigious math institute. At an “Ask Me Anything” event at Station F, he said, “We need more infrastructure, we need a European cloud, we need more European intensive computing centers, we need a European hardware industry, we need more European research centers — and that will take time and money. But it’s worth it because that will [help ensure] European sovereignty. And it should be constructed not in an ambiance of war but in the spirit of competition.”

Marc Schoenauer, 60, Villani’s guide in the AI netherworld, arrived at a dive bar south of Paris on a motorcycle, his long, wavy gray hair and beard askew, wearing a baggy sweater and jeans. If Villani’s attire and studied manner evoke a 19th century dandy with Goth overtones, Schoenauer gives off a former hippie vibe. A researcher at the French Institute for Research in Computer Science and Automation (Inria), Schoenauer has spent 30 years studying artificial intelligence and is the AI expert on Villani’s task force. He accepted the invitation, he said, “naively,” ignorant of the long hours and the months the report would go into overtime.

During the task force’s three-hour hearings, Schoenauer said, interviewees would, one by one, make a statement of their recommendations, which would be followed by a discussion, after which Villani would circle back to issues that interested him. Each task force member was responsible for a chapter of the approximately 200-page document, which would then be reread by their colleagues. In the final phase, a draft was distributed to French ministries for feedback on the feasibility of the recommendations and the necessary funding.

On working with Villani, Schoenauer commented, “He’s very impressive, first of all, because of the amount of work he can do.” He also noted that Villani had to head or contribute to other task forces and is a regular member of Parliament. Furthermore, Schoenauer said, “He remembers everything that was said in the hearings and fills up small notebooks. Later, he knows exactly which notebook he put the note in.”

“I think his motivation is pure in politics. It’s not at all to put himself forward,” said his task force collaborator Yann Bonnet, general secretary of the French digital advisory commission Conseil National du Numérique. “He’s a very good politician. … He always listened carefully, found compromises, was able to understand the balance of power. … It made it so everyone was happy to have contributed to the task force.”

“I’ve never seen him angry,” said another task force member, Anne-Charlotte Cornut. “I never heard him raise his voice in the six months we worked together. And we worked 24 hours a day.”

Schoenauer allowed that perhaps Villani enjoys his public profile and persuading others to his point of view. However, contrary to the “shocking” attitude of some cabinet members and politicians who “don’t give a shit,” he said of Villani, “I think he’s motivated to do something useful for France and mankind in general.”

Interviewed this month in a private room at Le Bourbon, a brasserie favored by the French political elite, Villani said he was motivated to enter politics because of what he saw as the recent “chaos” in French politics. And Macron’s pro-European, “neither right-wing nor left-wing” stance suited Villani’s own long-held political views.

“There was a need to serve the nation in a moment of great confusion, and [there was also] the idea that this was a chance that shouldn’t be missed.” He continued, sometimes spasmodically tapping the table for emphasis, “During the second round of voting [during the presidential election], it was possible that it was going to be a choice between the extreme left or the extreme right. It was chaos!”

Villani admits that, as a new representative, he had to learn “everything” about political life. “You learn the way laws are made, how political influence works, relations between [governmental] groups, cultural questions, work on constitutional reform,” he enumerated.

When it was suggested that some mathematicians might not understand the abandonment of research by one of its highest practitioners, he was slightly defensive. “Every time scientists have the feeling that there’s a subject that mixes science and politics — baf! — they come to see me, or they write me.” He said he had spoken to a half-dozen AI researchers just that morning who urgently wanted to explain their point of view. “I see in all these examples how important it is to have scientists in politics. It’s important for politics. It’s important for science.”

As for the long-awaited AI report, Villani is satisfied. “I think we have the diagnosis, I think we have the recommendations, and I think we’ve listened to enough people to be fairly sure of our recommendations,” he said, noting that not only had his team formally interviewed hundreds of experts, but his constant presence at conferences large and small, as well as his blanket coverage in the mainstream media have elicited feedback — good and not so good — from government and industry observers.

“When you do an interview in a mass circulation publication, if there’s something wrong in what you say, you can be sure that people will tell you so!” Villani remarked.

He paused to take a sip of orange juice. His pale skin seemed a bit gray, having lost the translucent glow observed during a previous meeting. While his colleague Marc Schoenauer was looking forward to returning to his research after the delivery of the AI report (“I learned a lot … but that’s it now”), Villani’s talents will be required for its implementation.

“It’s not the end, but it’s a step we’ve taken,” Villani said of the report’s completion. “We’re going to the next stage of the mission.” A waiter informed him he had a group of people waiting in the wings to meet with him. The deputé seemed game, if a little tired. He was still on the nation’s clock.