It’s hard to tell exactly when AI slop stopped being just a punchline involving Will Smith eating spaghetti and started feeling dangerous. That’s actually part of the problem.

This year, many of the popular AI tools that used to struggle with the number of fingers on a human hand gained the ability to generate photorealistic video — with audio — that costs users virtually nothing to create. Even trained scientists can’t always tell the difference between what’s real and what’s machine-made in digital form. But the most insidious applications of this remarkable technology are playing out on a much more personal level.

A few years ago, we started to hear more about voice cloning tools, which allow scammers to spoof a loved one’s voice in order to extort money from a victim. Large language models (LLMs), like ChatGPT, have also made any text-based scams more convincing and much cheaper to deploy. Imposter scams, which include romance scams, tech support fraud, and employment fraud, are on the rise, and thanks to increasingly high-quality AI, it’s harder than ever to know who’s real and who’s an imposter online.

When my daughter’s daycare recently sent us a picture of her entire class smiling in unison, my first thought was, “That’s cute!” My wife told me to look again and said, “That’s not our daughter.” She was right. All of the babies’ faces were AI-generated. 

That was the moment I knew that, despite my years as a tech reporter, constantly warning my boomer parents to take cybersecurity seriously, the scammers could get me. 

As AI gets better and cheaper, we’re all becoming easier targets. Your boomer parents are susceptible to getting scammed in the scary age of AI-powered fraud — but so is your Gen Z sibling, and so are you. 

The amount of money people over 60 have lost in scams skyrocketed from about $600 million in 2020 to about $3 billion in 2025, according to a new Federal Trade Commission report. The overall number of older adults who reported losing money to scammers, however, was actually 14 percent lower than the number of younger people who reported losses. Older adults are historically more likely to report being targeted but less likely to lose money to scammers, the FTC said. When they do lose money, they tend to lose a lot of it.

But younger generations are actually more likely to fall for certain scams, according to numerous surveys from the past five years. 

We should all expect to be bigger targets in the future, too, as scammers inundate us with cheaply produced but very realistic AI slop. “The sophistication of the types of scams that are happening to older adults has definitely moved forward,” Diana Freed, an assistant professor of data science and computer science at Brown, told me. “Having some type of family safety plan around these issues is very helpful.”

The two most important things to do in this stressful AI scam era: reduce the number of scammers that can get to you, and establish ways to verify that people — on the phone, online, in videos — are who they say they are. Assume that anything can be convincingly faked these days, and that your whole family will need help figuring out what’s real. 

Our best chance at getting through this tricky transition unscathed is to do it together.

Why you aren’t wrong to worry about your aging parents

It’s tempting to let generalizations guide your thinking when it comes to internet safety and behavior. Stubborn stereotypes tell us that older people are warm but incompetent, and that makes them more trusting, easy targets for scammers. We’ve all heard stories about someone’s grandma losing her life savings to criminals on the other side of the planet. Now we live in a world where AI can make those criminals sound like a grandchild on the phone, asking their grandparent to send money for an emergency. The same technology can make an aging parent sound like they’re asking their millennial son or daughter for urgent help, too.

“I think the security risks are certainly the same for younger people as well as older people,” said Ed Baig, who covers technology for AARP’s website and magazine.

Still, aging adults do have unique vulnerabilities that you should be aware of. 

As we get older, our brains start to change. Normal cognitive aging leads to declines in processing speed (so you’re a little bit slower), and memory (so you’re a little more forgetful). That doesn’t inherently make an older person more likely to be a scam victim, but some research suggests that low memory function makes people more susceptible to phishing scams. One 2012 study also linked less activity in the anterior insula (the part of the brain that processes disgust, among other things) in older subjects with more trustworthiness. There are other brain changes, like declines in analytical reasoning, that could explain why older adults tend to respond to certain scammer methods, like finding common ground or offering something free.

More than anything, though, older adults tend to have more money — or, at least, they tend to part with larger amounts of it. They are, in this sense, better marks. In 2024, according to the FTC, the median amount reported lost to fraud for people in their 30s was $450, but that number shoots up to $1,000 for people in their 70s and $1,650 for people in their 80s. 

In general, nobody likes to admit they fell for a scam. But the stakes are higher for some older adults, who might fear that they’ll lose their independence if they confess to their families that they’ve been a victim. They may also avoid reaching out to younger relatives to verify a shady-sounding exchange — or to ask questions in general — because they don’t want to bother anyone. 

None of this is taking the very real power of AI into the equation, by the way. Older adults use AI less often than younger generations. About one in four people over 65 have ever used a chatbot, according to Pew, although a separate survey found that over half of older adults have used voice-based assistants, like Alexa. (Thanks to recent upgrades, Alexa is gaining new AI powers, too.) 

So boomers may not be AI experts in large numbers, but they’re getting to know what it’s like to deal with chatbots. That still doesn’t mean they are fully aware of how much more believable scams have gotten thanks to AI. 

What to do with all your AI scam anxiety

Assuming we’re all equally doomed when it comes to AI scams — but also, equally capable of resisting — the best thing to do is prepare yourself and your loved ones. Here are some concrete steps you can take. 

Look at AI-generated content together to better understand the problem

Travis Bible, a Los Angeles-based filmmaker and editor in his 40s, told me he felt the need to alert his parents about potential scams after Google’s AI video generator got uncannily good. The third generation of the software, Veo 3, can turn text into photorealistic video with audio in a matter of moments, and Bible realized that the videos the tool can produce could fool even him. 

When simply describing the problem didn’t convince his parents to take it seriously, Bible got creative: He spent three days using AI to make a PSA warning of the dangers of deepfake scams and showed it to them. It worked. “I kind of expected them to be in disbelief, but honestly, they just kind of accepted it,” Bible told me. “Seeing the video made my warning click with them. Like, ‘Oh, it really is that good.’”

There are other PSAs like this floating around, but watching Bible’s made something click for me, too. I can’t really tell if half the clips are real or fake. You can look for tells, like glitches or visual typos, but AI image generators have gotten so good, it can be impossible to spot a fake at a glance. There are also AI tools that will tell you if a piece of media was generated by AI, like the Google-made SynthID Detector. But let’s be honest: When you’re in the heat of a scam, you don’t necessarily have the wherewithal to be testing the authenticity of stuff.

Talk about scams in a stigma-free environment

Older adults don’t want to be scam victims, but they also don’t want to be surveilled by their anxious relatives. To strike a balance, you can take the stigma out of potential victimhood by talking about how scams are creeping into your own life. Feel free to complain about the incessant string of scam texts about unpaid tolls you’ve been getting. If you’re using AI in your daily life, talk about it with your parents or grandparents, and you may learn more about what they’re doing too.

“Older people may turn to technology to ask questions to avoid burdening their children, meaning they may be hesitant to reach out to an adult child to ask for help with AI,” Robin Brewer, an associate professor of information at the University of Michigan, told me. Approaching these topics with generosity instead of judgment might mean your dad reaches out to you the next time he gets a weird text instead of clicking the link in the text to see where it takes him.

Create a family plan for verification

If everyone’s in agreement that we’re living through a scary time, you should create a family plan. This is useful for the older adults in your life, and it can be useful for you or your kids if you’re ever targeted. 

The basic idea is to create a chain of verification, so that if someone receives a call or an email — especially one claiming to be from a relative — they have another family member who can give it a second set of eyes. Scammers can easily spoof phone numbers, so it might look like the call is coming from your sister, when really it’s coming from Singapore. Make sure everyone in the family knows that urgency is a scammer’s best friend, and it’s always a good idea to slow down to try and verify. 

You might also want to agree on codewords to confirm that someone is who they say they are when, say, they call asking for “bail money.”

There’s also a cottage industry dedicated to selling you tech to protect your older loved ones and yourself from scams. You can pay for landline services from companies like Community Phone, Ooma, or teleCalm that protect against unwanted callers, filter, and offer caregivers some basic call monitoring features. Major carriers, like Verizon and AT&T, have filters and scam protection built into their services, and smartphones have features like call screening to help you avoid contact with a criminal. But ultimately nothing is 100 percent when it comes to stopping scammers from calling potential victims, since they’re constantly updating their tactics.

“Technology accelerating concentrates everything in our reality, and fraud will get even more powerful because the technology is getting better,” James Graham, Community Phone’s founder and CEO, told me. “And so you need technology that’s also getting better to prevent it, hopefully at a faster rate.”

So it’s a doom loop. Technology is helping criminals get better at stealing, so we need to invent new technology in order to thwart the thieves — until the next generation of tech comes along, at least. The truth is, we can’t stop scammers from assaulting our phone lines and inboxes any better than we can stop getting older with each passing day. 

We can, however, talk to each other about these constant threats. Warn your family and friends about the evolving threats that you’re seeing reported in the news or on social media. If your grandma sends you AI slop, talk to her about how AI slop gets made and how it can be dangerous. If your kid’s daycare sends you AI slop, make sure your spouse knows it’s AI slop, even if it’s cute.

I’m actually hopeful that things will get better. If you look back a couple of decades, when email was new, the volume of spam you’d get made it almost unusable. Spam filters got better, and we got better at spotting scams. Like the scammers themselves, we’ll have to keep adapting. But if we can teach an AI-generated Will Smith to eat spaghetti properly, as a society, we can do anything.

This is the free edition of Where Things Stand

Americans Voted Down Data Centers in Every Single Ballot Measure Except One

I’ve been chronicling the swelling backlash toward data center construction in communities across America — and how we thought it might translate into some losses for some Republicans in the midterms — in this evening newsletter for months, as regular readers will know.

The righteous outrage that Americans across the political spectrum feel toward Big Tech oligarchs coming in and draining crucial water supplies in their communities, polluting their air and driving up the costs of basic necessities — like electricity — was overwhelmingly apparent at the voting booth. Voters rejected data centers both literally and figuratively in dozens of races across the nation.

Let’s start with the literal backlash.

In a state of post-election bliss, John Nicely drove his daily school bus route on Wednesday morning through the tiny town of Perry, Ohio, a few miles south of Lake Erie.

“So, when are you taking down the signs, Mr. Nicely?” his students asked him as they filed inside, “because you kicked butt, buddy.” 

They were joking, of course, the way kids sometimes do with an adult they really like — and Nicely is, if my calls with him are any indication, a man who makes good on the promise of his name. But they were also right: He had kicked butt.

Of all the losers in the midterm elections perhaps none collapsed as decisively as data centers, those hulking, humming backbones of the AI boom, whose construction has incited a rage-fueled bipartisan backlash across the country. More than 40 data center-related measures made it to the ballot this Tuesday, and data centers lost all but one of them, often by dramatic margins. It is by far the most resounding referendum yet on the infrastructure of AI, a bipartisan revolt against a flagrantly opaque industry with a near-unanimous “no thanks,” or at the very least, a “probably not, but only if.” 

Even against the dramatic background of that national movement, Perry stands out. For much of the past year, the village has been embroiled in a bitter fight over plans for an enormous 230-acre data center — the equivalent of one-sixth of its surface area — quite literally in Nicely’s backyard, or at least in the center of the town he now calls home. But he and his fellow residents aren’t simply looking to stop the project from happening. 

While most of the measures that passed elsewhere outright ban data centers, or at least limit their construction, Perry’s Issue 16 does something more novel and maybe more consequential: It forces data centers to keep their promises. As crafted by Nicely and other advocates in Perry, it would require any data center in Perry that accepts a tax abatement or a subsidy — as almost all of them do — to take out a bond guaranteeing that even if the AI bubble bursts, even if companies fling all the data centers into space, or even if the facilities shrink to the size of a penny, the village will be left with more than empty promises and a gaping vacant campus. 

And on Tuesday, after months of sign-posting, petition drives, and unprecedentedly raucous village hall meetings, Nicely and his allies won. 

“I was absolutely choked up” as the results poured in, he told me on Wednesday, still a little bit teary over the final vote on Issue 16, which received a stunning 87 percent of Perry’s support. “I had the optimism that we would get it passed, but never by a landslide,” he said. “I couldn’t be prouder of this town than I am today.”

Data center companies should be paying attention. Combined with a blue tsunami that has carried in a slate of Democratic candidates who’ve campaigned on data center restrictions, and with more than half of voters calling AI a major issue in their choice for Congress, the industry cannot proceed with business as usual. As of now, if a data center wants to get built in the US, it will need to offer Americans a much better, more convincing deal. Issue 16 might show them the way.

Small town, big data center

Few towns know what a good deal looks like better than Perry, which has been molded for the past half-century by the looming cooling towers of its local nuclear power plant. 

When the Perry Nuclear Power Plant first opened in the 1980s, it brought hundreds of well-paying jobs and enough tax revenue to build luxury public school facilities like an Olympic-sized swimming pool. 

“I knew kids from high school who graduated and got these power plant jobs, and it was like hitting the lottery,” said Nicely, who grew up in the town next door. “They came in right off the bat to partner with the community,” bankrolling benefits such as garbage collection and free utilities for those most affected by “this thing being right in their backyards.”

Today, the station’s plumes of vapor are like pheromonal beacons, calling out to energy-hungry data centers. Earlier this year, the plant’s operator inked deals with Meta and the US Department of Energy to greatly expand its capacity, potentially making the facility into the largest of its kind in the country. But all that juice will do little good without the data centers that need it, and when a far-flung developer called the Province Group showed up in Perry in April, having already made a bunch of secret dealings with local policymakers behind closed doors, residents were livid. 

“I’m going, ‘Oh my God! They think that this town is nothing but a bunch of little hicks,’” said Nicely, who was recruited onto the village council a few months ago for his activism. He remembers thinking just that while he sat on his perch at the village hall, listening as the developer “tried to wow us with a bunch of spreadsheets,” rattling off promises for future jobs and tax revenue. Nicely, who is in his mid-60s, had a career working in business before he moved back to northeastern Ohio a few years ago to care for his mother — and, ultimately, drive the school bus, among other things. In his old life, he’d been a self-described “spin doctor,” and in the village hall he sniffed a lousy deal right away. If Perry’s nuclear plant once “truly took on a partnership role” when it came online, he said, then “these people came in with no regard to partner.”

Under the Province Group’s initial proposal, the data center would receive a 75 percent tax exemption for its first 15 years after construction. In return, once it started paying taxes, it promised to pay up plenty, to the tune of over $320 million over the next three decades, while ushering in the creation of hundreds of jobs. But 15 years is a long time to wait for a company to fork over the goods, and data centers are hardly known for going on long-term hiring sprees. 

Data centers, despite their odious reputation, can be good for local communities under the right conditions. Virginia’s Loudoun County has become well-known as “Data Center Alley” for its AI-fueled growth. There too, tax revenue from the industry has paid for fancy amenities, much like Perry’s nuclear plant did in the ’80s — indeed, Loudoun even got an Olympic pool of its own. And there are plenty of other proposals swirling for using the oodles of money behind data centers to improve local infrastructure. 

But Ohioans like Nicely had reason to fear they were about to get ripped off, partly because it had happened before. Back in 2019, Perry’s nuclear power plant — like other energy developments across the state — was on the verge of bankruptcy. Ohio passed HB 6, a last-ditch billion-dollar bailout funded by surcharges on Ohioans’ energy bills, in response. But while HB 6 saved the plant, it was later revealed to be part of the largest bribery scandal in Ohio history, one that sent the state’s former House Speaker Larry Householder to federal prison for two decades. 

What’s more, many of the policymakers involved in HB 6 have spent the past few years rolling out the red carpet for data center developers, enacting subsidies and tax breaks that have turned the state into one of the fastest-growing data center markets in the country. 

“They have no credibility on this issue,” said Nick Owens, a co-founder of Conserve Ohio, a group that sponsored ballot measures in 10 communities across the state. Owens is a former Republican candidate for state office who became an FBI informant in the HB 6 case.

After that scandal, “you would think there would be some muscle memory left in the system to realize that we shouldn’t be bought and paid for by all these special interests,” he told me. “But it’s still happening.” 

What if data centers had to keep their promises?

And that brings us to Issue 16, the ballot measure Nicely helped bring to Perry’s ballot on Election Day. Amid outrage from residents, the village council in Perry had already struck down the Province Group’s initial request for a tax abatement in July, putting the project in limbo. 

But thanks to the passage of Issue 16, if that developer or another like it tries to apply for one in Perry again, they will be forced to take out a kind of insurance on their promises, protecting the town in case the projected tax revenue never materializes as expected in the years that follow.

“That’s the threat with data centers, that in a couple of years, these projects will not be needed,” said Kasia Tarczynska, a senior research analyst at Good Jobs First, a group that studies government subsidies. “It is totally fair to say that if you want us to give you something, give us something in return and guarantee that you really want to be here.” 

“Initiatives like this really show that people are innovative and smart — they understand what they’re giving up. They understand that these corporations often come in, promise, and never deliver.”

Kasia Tarczynska, Good Jobs First senior research analyst

Many small towns, particularly across the Rust Belt, are acutely sensitive to the risks of an industry swooping in for a few sweet decades before disappearing, leaving decaying facilities and a local economy in turmoil. In recent years, states like Georgia and Texas have required projects to take out bonds similar to Perry’s to guarantee that local benefits actually materialize for developments that receive a large subsidy or tax break from the government, or to compel them to clean up after themselves if a project goes bust. 

“Initiatives like this really show that people are innovative and smart — they understand what they’re giving up,” Tarczynska said. “They understand that these corporations often come in, promise, and never deliver.”

If data centers want to continue to build in small towns across America, then they might need to become much more comfortable backing up their promises before they break ground. As Sam Nelson, a professor of political science at the University of Toledo, put it, “communities might be a lot more willing to be open to this development if they were actually going to see the money flowing.”

Nicely has been hearing from local policymakers across the state and the country interested in building out similar bonds into their data center deals, some of which have been inspired by Issue 16 in Perry. Even before the election, developers around the country had begun offering eye-popping and legally binding community benefits agreements promising to pump tens of millions of dollars into local schools and infrastructure. Those concessions may not assuage all the concerns that Americans have about data centers — which are, after all, arguably proxies for other fears about the madcap proliferation of AI. But they can still be meaningful, especially in communities that otherwise have the least to start with or the most to lose.

And Nicely’s town could certainly use the cash.

“Perry needs the money,” he told me. The town’s newest fire truck is now a decade old, and its ambulances are “in dire need of replacement,” he said. “There’s just no money to do it. So we have got to have some type of economic development.”

And that will probably require doing something with the sprawling property in the middle of town. It need not be a data center, Nicely said. But if it is, the town now has a lot of leverage to go with all the power flowing from its plant. “If you want to be business partners then come in as business partners. Offer the people something up front, and tell them how you’re going to invest in their community, not how it’s going to benefit you,” he said. “Or go away.”

Google is tying up AI Overviews with traditional search, angering parents and educators who worry that easy, machine-generated answers are leaving kids less capable.

This week on Uncanny Valley, we also discuss Anthropic’s IPO filing and the volunteers gearing up to counter ICE at the midterms.

Big Tech is pouring billions into AI infrastructure, but mounting debt, negative cash flow, and growing investor skepticism are raising questions about the sustainability of the AI spending boom.

Six weeks after the EU unveiled its much-anticipated proposal to protect kids across the bloc from social media, European governments are pressing ahead with their own rules. 

National political leaders had spent months asking for unified age restrictions for social media access across the European Union, to avoid 27 different sets of rules. European Commission President Ursula von der Leyen’s response came in September with the EU Kids Act, a signature initiative of her second term that would ban under-13s from accessing most current social media platforms and severely restrict access for under-15s. 

But it is becoming clear that individual countries don’t want to wait for the Brussels legislative wheels to turn — and in some cases, are proceeding with national laws that would kick even more kids off the platforms than what von der Leyen has in mind.  

“While we wait for the EU Kids Act – we are pushing forward with national legislation,” Denmark’s digital minister Christina Egelund said. “The problems are urgent. Children and parents should not be asked to wait.” 

Last week, Greece’s parliament passed a bill to bar social media access for under-15s, while Italy’s own stalled efforts to shield under-14s suddenly found new winds.  

By Thursday, the Commission will have to give the first indication of how it will deal with national laws. Its response to Austria’s proposal to bar social media access for under-14s will be the first test case.

Capitals continue to ultimately favor a unified EU law, which is in line with existing rules that give Brussels oversight of big platforms like Instagram and TikTok. A harmonized law also avoids fragmenting the market with different rules in different countries.  

Yet agreeing on a final text for the legislation and implementing the Kids Act will take more than a year — too long for national leaders eager to show they are serious about protecting the mental health of kids and reining in Big Tech. The age at which limits are set is also a sticking point, with countries preferring various thresholds between 13 and 16.  

“Greece decided to move forward with its own legislation on minors’ access to social media because protecting children is a pressing issue that calls for timely and concrete action,” a spokesperson for the country’s digital ministry said. “We have a responsibility to address an issue that is already affecting the daily lives of children and their families,” they said, adding that the Greek law is “complementary” to the EU’s and that Athens is still in favor of a bloc-wide solution.  

Portugal formally notified the Commission of its own law a little over a week after the Kids Act was proposed, formally asking the EU executive to say if Lisbon’s plans are in line with European rules. Cyprus’s ministers approved their own ban for under-15s on the day Brussels proposed its law. Austria and Germany have also not shown any signs of slowing down. Many of these countries had backed an EU-wide solution.  

The current trajectory could create a confusing situation for minors, who may find themselves locked out of social media in January 2027 only to then regain access under parental controls once the Kids Act is implemented later that year or in 2028. 

The Commission had stressed that the law would bring about “maximum harmonization,” meaning national laws would be scrapped if and when the EU-wide measures came into force.  

When France first formally informed the Commission of its national law to ban social media access for under-15s, “we said very clearly: you should be able, and you should be ready to repeal this law once there are European measures in place,” a senior Commission official told reporters when announcing the law in September.  

But a few weeks later, Commission spokesperson Thomas Regnier told POLITICO that the EU executive was still considering how to respond to the laws that have been submitted to be checked for compatibility with European rules.  

The Kids Act will have to be negotiated between the Commission, the European Parliament and EU governments, meaning it may not be finalized for months. It would then come into force six months later. The senior official estimated negotiations would take eight to 12 months starting next September, putting the implementation date well into 2028.  

Capitals had backed a pan-European solution in part because it allows enforcement to align with the landmark Digital Services Act, which gives the power to oversee big platforms like Instagram and TikTok to Brussels. Industry and some civil society groups were also supportive of an EU-wide solution because they wouldn’t have to deal with dozens of national laws and different ages in different countries.  

But public sentiment around social media and teen mental health has taken a severe turn for the worse in recent years, culminating in multimillion- and, in some cases, billion-dollar settlements against Meta and TikTok across the U.S.  

In the meantime, European leaders and von der Leyen have spent the last year promising the public age restrictions for social media.  

“A European age limit will send a much stronger signal to Big Tech – enough is enough,” said Denmark’s Egelund. “It will also place the responsibility firmly and directly on those who should bear it – the tech giants.”  

This week on “Uncanny Valley,” we go behind the scenes of our Nathan Fielder cover story, unpack tech billionaires’ generous promises, and discuss the suspicious letters sent to US election offices.

An earnings briefing on Wednesday revealed ad revenue has slowed since the European Union introduced the Less-Personalized Ads model at the beginning of this year.

Meta’s board feared that its CEO’s departure could disrupt the company’s AI ambitions. A new stock agreement gives him financial incentives to stay put.

BRUSSELS — Europe’s biggest media companies have asked regulators in Brussels and London to stop Meta from introducing new terms they say would let it copy publishers’ websites to build artificial intelligence, as a backlash that began in the U.S. spreads across Europe.

In a letter Monday, the European Publishers Council, whose members include POLITICO’s owner Axel Springer, called on EU competition boss Teresa Ribera and tech chief Henna Virkkunen to press Meta to put the terms on hold before they take effect Oct. 30.

The group wants Brussels to check them against EU copyright and competition law, and has requested an urgent meeting with the teams of both commissioners.

“The Commission remains committed to ensure that Meta’s practices fully comply with all applicable obligations under EU law, including competition rules and the Digital Markets Act,” Commission spokesperson Ricardo Cardoso told POLITICO, referring to the EU’s rulebook for Big Tech. He said Brussels was aware of both Meta’s changes and publishers’ concerns.

The U.K. competition watchdog has moved as well. British publishers have complained to the Competition and Markets Authority that Meta is so dominant they are compelled to accept its terms. 

The watchdog’s chief, Sarah Cardell, told British lawmakers on Tuesday that it had asked Meta to explain the changes and would meet the company on Wednesday. She named Meta as a possible candidate for Britain’s new rules for tech giants.

The terms cover businesses that use Meta’s tools to run Facebook pages, Instagram accounts or ads, which includes almost every newsroom. 

The fight began in the U.S. last week, when Nieman Lab, a journalism research outlet at Harvard, reported on the terms after receiving a tip from a small business owner.

As first written, they let Meta send automated programs to copy websites linked to those accounts and use the material to “train” its products, Nieman Lab concluded. There is no opt-out, and a business that continues to use Meta’s tools after Oct. 30 will be deemed to have consented.

Publishers say the terms would hand Meta content for free that other AI companies pay to license.

Rapid reaction

On Friday, Meta communications chief Andy Stone responded on X that the Nieman Lab report was “false” and that there was “no change to how Meta uses web crawlers.”

Stone said the update aimed to let Meta pull accurate product information from shopping sites for ads, not to scrape news sites. The company also said the crawling language already existed in other Meta terms and was just being gathered in one place.

Late Friday, however, Meta revised the terms and deleted the word “train,” saying the terms cover only its advertising and business products.

“We’ve heard you and want to clarify,” a spokesperson for Meta told POLITICO over the weekend, adding the change was “not about scraping news sites.”

Publishers say that changes little, however. 

Meta has given them verbal assurances that they can keep using its tools without surrendering their content to AI, but the terms themselves don’t guarantee that, the European Publishers Council’s letter says.

“Meta’s supposed Friday night correction is a farce,” said Jason Kint, who heads Digital Content Next, a U.S. publisher group whose members include Axel Springer. “Exactly what you’d expect from a gatekeeper notorious for abusing its market power.”

Others have gone further. 

“This is unscrupulous theft,” Germany’s newspaper and magazine associations said in a joint statement on Tuesday. They want Brussels to block the terms while it reviews them, warning that once articles are copied into AI systems, they can’t be clawed back.

Joseph Bambridge in London contributed reporting.

TikTok and other sites were told to curb protest footage and AI-altered images, while online retailers like Vinted and eBay reportedly signed an agreement stop looted goods from flooding resale sites.