Business

The Race to Prevent ‘the Worst Case Scenario for Machine Learning’

Published

2 years ago

June 24, 2023

The Race to Prevent ‘the Worst Case Scenario for Machine Learning’

Dave Willner has had a front-row seat to the evolution of the worst things on the internet.

He started working at Facebook in 2008, back when social media companies were making up their rules as they went along. As the company’s head of content policy, it was Mr. Willner who wrote Facebook’s first official community standards more than a decade ago, turning what he has said was an informal one-page list that mostly boiled down to a ban on “Hitler and naked people” into what is now a voluminous catalog of slurs, crimes and other grotesqueries that are banned across all of Meta’s platforms.

So last year, when the San Francisco artificial intelligence lab OpenAI was preparing to launch Dall-E, a tool that allows anyone to instantly create an image by describing it in a few words, the company tapped Mr. Willner to be its head of trust and safety. Initially, that meant sifting through all of the images and prompts that Dall-E’s filters flagged as potential violations — and figuring out ways to prevent would-be violators from succeeding.

It didn’t take long in the job before Mr. Willner found himself considering a familiar threat.

Just as child predators had for years used Facebook and other major tech platforms to disseminate pictures of child sexual abuse, they were now attempting to use Dall-E to create entirely new ones. “I am not surprised that it was a thing that people would attempt to do,” Mr. Willner said. “But to be very clear, neither were the folks at OpenAI.”

For all of the recent talk of the hypothetical existential risks of generative A.I., experts say it is this immediate threat — child predators using new A.I. tools already — that deserves the industry’s undivided attention.

In a newly published paper by the Stanford Internet Observatory and Thorn, a nonprofit that fights the spread of child sexual abuse online, researchers found that, since last August, there has been a small but meaningful uptick in the amount of photorealistic A.I.-generated child sexual abuse material circulating on the dark web.

According to Thorn’s researchers, this has manifested for the most part in imagery that uses the likeness of real victims but visualizes them in new poses, being subjected to new and increasingly egregious forms of sexual violence. The majority of these images, the researchers found, have been generated not by Dall-E but by open-source tools that were developed and released with few protections in place.

In their paper, the researchers reported that less than 1 percent of child sexual abuse material found in a sample of known predatory communities appeared to be photorealistic A.I.-generated images. But given the breakneck pace of development of these generative A.I. tools, the researchers predict that number will only grow.

“Within a year, we’re going to be reaching very much a problem state in this area,” said David Thiel, the chief technologist of the Stanford Internet Observatory, who co-wrote the paper with Thorn’s director of data science, Dr. Rebecca Portnoff, and Thorn’s head of research, Melissa Stroebel. “This is absolutely the worst case scenario for machine learning that I can think of.”

Dr. Portnoff has been working on machine learning and child safety for more than a decade.

To her, the idea that a company like OpenAI is already thinking about this issue speaks to the fact that this field is at least on a faster learning curve than the social media giants were in their earliest days.

“The posture is different today,” said Dr. Portnoff.

Still, she said, “If I could rewind the clock, it would be a year ago.”

‘We trust people’

In 2003, Congress passed a law banning “computer-generated child pornography” — a rare instance of congressional future-proofing. But at the time, creating such images was both prohibitively expensive and technically complex.

The cost and complexity of creating these images has been steadily declining, but changed last August with the public debut of Stable Diffusion, a free, open-source text-to-image generator developed by Stability AI, a machine learning company based in London.

In its earliest iteration, Stable Diffusion placed few limits on the kind of images its model could produce, including ones containing nudity. “We trust people, and we trust the community,” the company’s chief executive, Emad Mostaque, told The New York Times last fall.

In a statement, Motez Bishara, the director of communications for Stability AI, said that the company prohibited misuse of its technology for “illegal or immoral” purposes, including the creation of child sexual abuse material. “We strongly support law enforcement efforts against those who misuse our products for illegal or nefarious purposes,” Mr. Bishara said.

Because the model is open-source, developers can download and modify the code on their own computers and use it to generate, among other things, realistic adult pornography. In their paper, the researchers at Thorn and the Stanford Internet Observatory found that predators have tweaked those models so that they are capable of creating sexually explicit images of children, too. The researchers demonstrate a sanitized version of this in the report, by modifying one A.I.-generated image of a woman until it looks like an image of Audrey Hepburn as a child.

Stability AI has since released filters that try to block what the company calls “unsafe and inappropriate content.” And newer versions of the technology were built using data sets that exclude content deemed “not safe for work.” But, according to Mr. Thiel, people are still using the older model to produce imagery that the newer one prohibits.

Unlike Stable Diffusion, Dall-E is not open-source and is only accessible through OpenAI’s own interface. The model was also developed with many more safeguards in place to prohibit the creation of even legal nude imagery of adults. “The models themselves have a tendency to refuse to have sexual conversations with you,” Mr. Willner said. “We do that mostly out of prudence around some of these darker sexual topics.”

The company also implemented guardrails early on to prevent people from using certain words or phrases in their Dall-E prompts. But Mr. Willner said predators still try to game the system by using what researchers call “visual synonyms” — creative terms to evade guardrails while describing the images they want to produce.

“If you remove the model’s knowledge of what blood looks like, it still knows what water looks like, and it knows what the color red is,” Mr. Willner said. “That problem also exists for sexual content.”

Thorn has a tool called Safer, which scans images for child abuse and helps companies report them to the National Center for Missing and Exploited Children, which runs a federally designated clearinghouse of suspected child sexual abuse material. OpenAI uses Safer to scan content that people upload to Dall-E’s editing tool. That’s useful for catching real images of children, but Mr. Willner said that even the most sophisticated automated tools could struggle to accurately identify A.I.-generated imagery.

That is an emerging concern among child safety experts: That A.I. will not just be used to create new images of real children but also to make explicit imagery of children who do not exist.

That content is illegal on its own and will need to be reported. But this possibility has also led to concerns that the federal clearinghouse may become further inundated with fake imagery that would complicate efforts to identify real victims. Last year alone, the center’s CyberTipline received roughly 32 million reports.

“If we start receiving reports, will we be able to know? Will they be tagged or be able to be differentiated from images of real children?” said Yiota Souras, the general counsel of the National Center for Missing and Exploited Children.

At least some of those answers will need to come not just from A.I. companies, like OpenAI and Stability AI, but from companies that run messaging apps or social media platforms, like Meta, which is the top reporter to the CyberTipline.

Last year, more than 27 million tips came from Facebook, WhatsApp and Instagram alone. Already, tech companies use a classification system, developed by an industry alliance called the Tech Coalition, to categorize suspected child sexual abuse material by the victim’s apparent age and the nature of the acts depicted. In their paper, the Thorn and Stanford researchers argue that these classifications should be broadened to also reflect whether an image was computer-generated.

In a statement to The New York Times, Meta’s global head of safety, Antigone Davis, said, “We’re working to be purposeful and evidence-based in our approach to A.I.-generated content, like understanding when the inclusion of identifying information would be most beneficial and how that information should be conveyed.” Ms. Davis said the company would be working with the National Center for Missing and Exploited Children to determine the best way forward.

Beyond the responsibilities of platforms, researchers argue that there is more that A.I. companies themselves can be doing. Specifically, they could train their models to not create images of child nudity and to clearly identify images as generated by artificial intelligence as they make their way around the internet. This would mean baking a watermark into those images that is more difficult to remove than the ones either Stability AI or OpenAI have already implemented.

As lawmakers look to regulate A.I., experts view mandating some form of watermarking or provenance tracing as key to fighting not only child sexual abuse material but also misinformation.

“You’re only as good as the lowest common denominator here, which is why you want a regulatory regime,” said Hany Farid, a professor of digital forensics at the University of California, Berkeley.

Professor Farid is responsible for developing PhotoDNA, a tool launched in 2009 by Microsoft, which many tech companies now use to automatically find and block known child sexual abuse imagery. Mr. Farid said tech giants were too slow to implement that technology after it was developed, enabling the scourge of child sexual abuse material to openly fester for years. He is currently working with a number of tech companies to create a new technical standard for tracing A.I.-generated imagery. Stability AI is among the companies planning to implement this standard.

Another open question is how the court system will treat cases brought against creators of A.I.-generated child sexual abuse material — and what liability A.I. companies will have. Though the law against “computer-generated child pornography” has been on the books for two decades, it’s never been tested in court. An earlier law that tried to ban what was then referred to as virtual child pornography was struck down by the Supreme Court in 2002 for infringing on speech.

Members of the European Commission, the White House and the U.S. Senate Judiciary Committee have been briefed on Stanford and Thorn’s findings. It is critical, Mr. Thiel said, that companies and lawmakers find answers to these questions before the technology advances even further to include things like full motion video. “We’ve got to get it before then,” Mr. Thiel said.

Julie Cordua, the chief executive of Thorn, said the researchers’ findings should be seen as a warning — and an opportunity. Unlike the social media giants who woke up to the ways their platforms were enabling child predators years too late, Ms. Cordua argues, there’s still time to prevent the problem of AI-generated child abuse from spiraling out of control.

“We know what these companies should be doing,” Ms. Cordua said. “We just need to do it.”

Business

Help! I Couldn’t Take My Tall-Ship Voyage, and I Want My Money Back.

Published

5 hours ago

March 6, 2025

Press Room

Help! I Couldn’t Take My Tall-Ship Voyage, and I Want My Money Back.

Dear Tripped Up,

Last summer, I booked a five-day sailing trip with Tall Ship Experience, a company based in Spain. For 1,350 euros, or $1,450, I would be a volunteer on the crew of the Atlantis, sailing between two ports in Italy. But eight days before, I had a bad fall that resulted in multiple injuries, including eight stitches to my face that doctors said I could not expose to sun or water. The Tall Ship Experience website clearly states that I could cancel for a full refund up to seven days before the trip. But the company revealed it was just an intermediary and the Dutch organization actually running the trip, Tallship Company, had different rules, under which I was refunded 10 percent. I offered to take credit for a future trip, to no avail. Finally, I disputed the charges with my credit card issuer, American Express. But Tall Ship Experience provided a completely different set of terms to Amex, saying I canceled one day in advance. The charges were reinstated. Can you help? Martha, Los Angeles

Dear Martha,

This story reads like a greatest-hits playlist of travel industry traps: a middleman shirking responsibility, terms and conditions run amok, a credit card chargeback gone wrong, and the maddening barriers to pursuing justice against a foreign company. However, the documentation you sent was so complete and the company’s website so confusing that I was sure Tall Ship Experience would quickly refund you.

Tallship Company did not respond to requests for comments, but did nothing wrong. It simply followed its own terms and conditions that Tall Ship Experience, as a middleman, should have made clear to you. When you canceled, Tallship Company sent back a 10 percent refund to Tall Ship Experience to then send to you.

That’s why I was surprised that the stubborn (though exceedingly polite) Tall Ship Experience spokeswoman who responded to me on behalf of the Seville-based organization argued repeatedly that although she regretted your disappointment, Tall Ship Experience was not at fault. At one point she suggested you should have purchased travel insurance, even as the company scrambled to adjust and update its website as we emailed.

Before the changes, the site contained two distinct and contradictory sets of terms and conditions: one for customers who purchased via the website’s English and French versions, and another on the Spanish version. (Confusingly, both documents were in Spanish.)

The English/French version — the one you had seen — promised customers a full refund for trips canceled more than seven days in advance. The Spanish one is vastly more complex, offering distinct cancellation terms for each ship. The Atlantis offered customers in your situation only 10 percent back.

Enter the stubborn spokeswoman: “The terms and conditions in Spanish correctly reflected the cancellation policy of the ship in the moment the client made the reservation,” she wrote via email. “We are conscious that at the time, the English version of the terms was not updated, which may have generated confusion. However, the official terms of the reservation were applied correctly.”

In other words, customers should somehow know to ignore one contract and seek out another on a different part of the site, both in a language they may not read.

But I am no expert in Spanish consumer law, so I got in touch with two people who are: Marta Valls Sierra, head of the consumer rights practice at Marimón Abogados, a law firm based in Barcelona; and Fernando Peña López, a professor at the Universidade da Coruña in A Coruña.

They examined the documentation and each concluded independently that Tall Ship Experience had violated basic Spanish consumer statutes. When I passed along their convincing points to the spokeswoman and alerted her that you were considering taking the company to Spanish small-claims court, she finally said it would refund you the remaining €1,215.

I felt a bit sheepish about exerting so much pressure on this small company — actually, an arm of the nonprofit Nao Victoria Foundation, which operates several replicas of historic ships — but the company should have taken much more care when it set up its website, Ms. Valls Sierra told me.

“If in your terms and conditions you say that up until seven days before departure you have the right to cancel,” she said in an interview, “and a consumer comes and says, ‘I want to cancel,’ you have to cancel their trip and return their money. They can’t use ‘Sorry, we forgot to put it on one web page, but we put it on another web page’ as an excuse.”

It is a principle of consumer law, she added, that confusing or contradictory contracts are interpreted in favor of the consumer.

The other troubling issue with the website is that you had no way of knowing that your trip was not operated by Tall Ship Experience. There was no such mention I could find on the website, which relies on marketing copy like this: “On board you will learn everything you need to know that will allow you to become one of our crew.”

Dr. Peña López, the law professor, wrote me in an email that “Tall Ship Experience is obligated to inform the consumer about the service it provides in an accessible and understandable manner, clearly indicating whether it is an intermediary.” He added that Tall Ship Experience “clearly” presented itself as the ship’s operator in this case.

As I mentioned, Tall Ship Experience did begin updating its site almost as soon as I got in touch, calling itself a “marketplace” for experiences and posting the correct terms and conditions (in the correct languages) on its English and French pages.

But Tall Ship Experience agreed to a refund only after I sent the company a compilation of the two experts’ legal analyses. “We are dedicated to creating experiences aboard unique boats, and not to legal matters,” came the spokeswoman’s response. “Regardless of which party is correct in this case, we would like to refund the full amount. We look forward to putting this to rest and to focus on continuing to improve customer experiences.”

You also said that American Express had let you down, by taking the company’s word over yours when you contested the charge. It is true that the document Tall Ship Experience sent to Amex (which forwarded it to you, who forwarded it to me), is wildly inaccurate, including only the terms favorable to the company and saying you canceled only one day in advance.

A spokeswoman for American Express emailed me a statement saying that the company “takes into account both the card member and the merchant perspectives.” But travelers should not mistake credit card issuers for crack investigators who will leave no stone unturned in pursuit of travel justice. A chargeback request works best when the problem is straightforward — you were charged more than you agreed to pay, or you never agreed to pay at all. Asking your card issuer to do a deep dive into terms and conditions is a much longer shot.

And as we’ve seen before (and might be seeing in this case) such chargeback requests often anger the companies involved to the point that they refuse to deal with you further.

If all else had failed, as I told you before the company gave in, you could have requested a “juicio verbal,” Spain’s version of a small-claims-court proceeding, via videoconference. It would not have been easy, said Dr. Peña López. Cases under €2,000 do not require a lawyer, but they do require you to have a Foreigner Identification Number, to fill out forms in legal Spanish (A.I. might help) and to find an interpreter to be by your side.

When I finally told you — in our 39th email! — you’d get a refund, you told me you had been “almost looking forward to a Spanish small-claims experience.” I admire your spirit, although I suspect it would have been quickly broken by bureaucratic and linguistic barriers.

If you need advice about a best-laid travel plan that went awry, send an email to TrippedUp@nytimes.com.

Follow New York Times Travel on Instagram and sign up for our Travel Dispatch newsletter to get expert tips on traveling smarter and inspiration for your next vacation. Dreaming up a future getaway or just armchair traveling? Check out our 52 Places to Go in 2025.

Business

In dizzying reversal, Trump pauses tariffs on most Mexican products

Published

6 hours ago

March 6, 2025

Press Room

In dizzying reversal, Trump pauses tariffs on most Mexican products

MEXICO CITY — In a dizzying turn, President Trump said Thursday that the U.S. would temporarily reverse the sweeping tariffs it imposed just days ago on most Mexican products.

In a post on Truth Social, Trump said he would delay for one month the imposition of 25% taxes on Mexican imports that fall under a free trade agreement that he negotiated during his last term.

His remarks follow comments from U.S. Commerce Secretary Howard Lutnick, who on Thursday said in a television interview that Trump was “likely” to temporarily suspend 25% tariffs on Canada and Mexico for most products and services, widening an exemption that was granted Wednesday only to vehicles.

Lutnick told CNBC that the one-month delay in the import taxes “will likely cover all USMCA-compliant goods and services,” a reference to the U.S.-Mexico-Canada trade agreement, the North America free trade pact Trump negotiated in his last term. Lutnick said around half of what the U.S. imports from Mexico and Canada would be eligible.

Lutnick said the reprieve will last only until April 2, when the Trump administration has said it will impose reciprocal tariffs on countries to match the ones they have on U.S. exports. Later, he said that if Canada and Mexico don’t do enough to stop fentanyl from entering the United States, the 25% tariffs could be reapplied in a month as well.

On Tuesday, the U.S. began placing duties of 25% on imported goods from Mexico and Canada, with a 10% rate on Canadian energy products. It also began imposing a new 10% tax on all imports from China.

Trump has said the tariffs are punishment because the three countries haven’t done enough to stop the flow of immigrants without proper documentation and drugs into the United States — and are an attempt to lure manufacturing back to the United States.

China and Canada responded forcefully, both imposing retaliatory tariffs on U.S. goods. Mexican President Claudia Sheinbaum had said that Mexico would also respond with counter tariffs, and had planned to announce them Sunday at a public rally in Mexico City’s central square.

In Canada, Prime Minister Justin Trudeau said he welcomed news that the U.S. would delay, but said Canada’s imposition of retaliatory tariffs will remain in place for now. “We will not be backing down from our response tariffs until such a time as the unjustified American tariffs [on] Canadian goods are lifted,” he said.

Trudeau told reporters that the U.S. and Canada are “actively engaged in ongoing conversations in trying to make sure these tariffs don’t overly harm” certain sectors and workers.

Business

Trump’s Cuts to Federal Work Force Push Out Young Employees

Published

17 hours ago

March 6, 2025

Press Room

Trump’s Cuts to Federal Work Force Push Out Young Employees

About six months ago, Alex Brunet, a recent Northwestern University graduate, moved to Washington and started a new job at the Consumer Financial Protection Bureau as an honors paralegal. It was fitting for Mr. Brunet, 23, who said he had wanted to work in public service for as long as he could remember and help “craft an economy that works better for everyone.”

But about 15 minutes before he was going to head to dinner with his girlfriend on the night before Valentine’s Day, an email landed in his inbox informing him that he would be terminated by the end of the day — making him one of many young workers who have been caught up in the Trump administration’s rapid wave of firings.

“It’s discouraging to all of us,” Mr. Brunet said. “We’ve lost, for now at least, the opportunity to do something that matters.”

Among the federal workers whose careers and lives have been upended in recent weeks are those who represent the next generation of civil servants and are now wrestling with whether they can even consider a future in public service.

The Trump administration’s moves to reduce the size of the bureaucracy have had an outsize impact on these early career workers. Many of them were probationary employees who were in their roles for less than one or two years, and were among the first to be targeted for termination. The administration also ended the Presidential Management Fellows Program, a prestigious two-year training program for recent graduates interested in civil service, and canceled entry-level job offers.

The firings of young people across the government could have a long-term effect on the ability to replenish the bureaucracy with those who have cutting-edge skills and knowledge, experts warn. Donald F. Kettl, a former dean in the School of Public Policy at the University of Maryland, says that young workers bring skills “the government needs” in fields like information technology, medicine and environmental protection.

“What I am very afraid of is that we will lose an entire generation of younger workers who are either highly trained or would have been highly trained and equipped to help the government,” Mr. Kettl said. “The implications are huge.”

The administration’s downsizing could have a lasting impact, deterring young workers from joining the ranks of the federal government for years, Mr. Kettl said.

About 34 percent of federal workers who have been in their roles for less than a year are under the age of 30, according to data from the Office of Personnel Management. The largest single category of federal workers with less than a year of service are 25- to 29-year-olds.

The federal government already has an “underlying problem” recruiting and retaining young workers, said Max Stier, the president of the Partnership for Public Service. Only about 9 percent of the 2.3 million federal workers are under the age of 30.

“They’re going after what may be easiest to get rid of rather than what is actually going to make our government more efficient,” Mr. Stier said.

Trump administration officials and the billionaire Elon Musk, whom the president has tasked with shrinking the federal government, have defended their efforts to cut the work force.

“President Trump returned to Washington with a mandate from the American people to bring about unprecedented change in our federal government to uproot waste, fraud and abuse,” Harrison Fields, a White House spokesman, said in a statement.

Mr. Trump has vowed to make large-scale reductions to the work force, swiftly pushing through drastic changes that have hit some roadblocks in court.

Last week, a federal judge determined that directives sent to agencies by the Office of Personnel Management calling for probationary employees to be terminated were illegal, and the agency has since revised its guidance. Still it is unclear how many workers could be reinstated.

The abrupt firings that have played out across the government so far came as a shock to young employees.

They described being sent curt messages about their terminations that cited claims about their performance they said were unjustified. There was a frantic scramble to download performance reviews and tax documents before they were locked out of systems. Some said they had to notify their direct supervisors themselves that they had just been fired.

On the morning of Feb. 17, Alexander Hymowitz sat down to check his email when he saw a message that arrived in his inbox at 9:45 p.m. the night before. An attached letter said that he had not yet finished his trial period and was being terminated from his position as a presidential management fellow at the Agriculture Department. It also said that the agency determined, based on his performance, that he had not demonstrated that his “further employment at the agency would be in the public interest.”

Mr. Hymowitz, 29, said he was dumbfounded. “My initial thought was, obviously something is wrong,” he said. “How could I get terminated for performance when I’ve never had a performance review?”

Mr. Hymowitz, who had worked on antitrust cases and investigations in the poultry and cattle markets for about six months, said he was not given many further instructions. The next day, he decided to walk into the office and drop off his work equipment. “I just assumed that’s what people do when they get fired,” he said.

Around 8 p.m. on Feb. 11, Nicole Cabañez, an honors attorney at the Consumer Financial Protection Bureau, found out that she had been terminated after she realized she could not log into her work laptop. Ms. Cabañez, 30, worked in the agency’s enforcement division for about four months, investigating companies that violated consumer financial laws.

“I was prepared to help make the world better,” Ms. Cabañez said. “It’s honestly very disappointing that I never got that chance.”

During her first year at Yale Law School, Ms. Cabañez said she originally planned to work at a large law firm, where she would have defended companies and made a lucrative income after graduation. But she said she wanted to work in public service to help people get relief through the legal system.

Ms. Cabañez said she was now applying for jobs with nonprofits, public interest law firms and local governments. But she said she worried that the job market, especially in Washington, would be “flooded with public servants.” She said she could not file for unemployment benefits for three weeks because her agency had not sent her all of the necessary documents until recently.

The impacts have stretched beyond Washington, reaching federal workers across the country, including in Republican-led states.

At 3:55 p.m. on Feb. 13, Ashlyn Naylor, a permanent seasonal technician for the U.S. Forest Service in Chatsworth, Ga., received a call from one of her supervisors who informed her that she would be fired after working there for about nine months. Ms. Naylor said she initially wanted to stay at the agency for the rest of her career.

“It was where I have wanted to be for so long, and it was everything that I expected it to be from Day 1,” Ms. Naylor said.

Ms. Naylor, 24, said she felt a mixture of anger and disbelief. She said her performance evaluations showed she was an “excellent worker,” and she did not understand why she was fired. Although she said she was devastated to lose her job, which primarily involved clearing walking trails in the Chattahoochee-Oconee National Forest, she was not sure if she would return to the agency in the future.

“It would be really hard to trust the federal government if I were to go back,” Ms. Naylor said. She said she was considering enrolling in trade school and possibly becoming a welder since she is still “young enough” to easily change her career.

Although some said their experiences have discouraged them from pursuing jobs with the federal government again, some said they were intent on returning.

Jesus Murillo, 27, was fired on Valentine’s Day after about a year and a half working as a presidential management fellow at the Department of Housing and Urban Development, where he helped manage billions of dollars in economic development grants. After standing in countless food bank lines and working in fields picking walnuts to help his family earn additional income growing up, Mr. Murillo said he wanted to work in public service to aid the lowest income earners.

“I’ve put so much into this because I want to be a public leader to now figure out that my government tells me that my job is useless,” Mr. Murillo said. “I think that was just a smack in the face.”

Still, he said he would work for the federal government again.

“For us, it’s not a partisan thing,” Mr. Murillo said. “We’re there to carry out the mission, which is to be of service to the American public.”