Tag: ai

Anthropic Blocks Foreign Access To Fable 5, Mythos 5 After U.S. National Security Order

About four days after Anthropic released Claude Fable 5, a next-generation “Mythos-class” AI model, the frontier AI lab led by Dario Amodei revealed late Friday that it was disabling foreign customers’ access to these cutting-edge models, citing an export-control directive from the federal government.

“The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees,” Anthropic wrote on X around 9 p.m. ET.

The AI lab’s website stated that the federal directive was received around 5:21 p.m. ET. To ensure compliance, the lab was forced to shut off Fable 5 and Mythos 5 for all customers.

Anthropic continued, “The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance. Access to all other Claude models is not affected.”

Anthropic pointed out that it understands the government’s concern centers on a potential method of bypassing, or “jailbreaking,” Fable 5.

Dario’s company laid out some of Fable’s safeguards:

  • We have instituted strong safeguards that greatly reduce the likelihood that Fable is misused for tasks related to cybersecurity (among others). In fact, our safeguards are so strong that many users have complained that they are overly broad.
  • In the weeks leading up to the launch of Fable, Anthropic worked with the US government, the UK AISI, multiple private third-party organizations and internal teams to red-team Fable’s safeguards for thousands of hours in total.
  • These tests showed that Fable’s safeguards are substantially more effective than those of any previously deployed model.
  • No testers have yet been able to find a universal jailbreak—a jailbreak method that can very broadly bypass the model’s safeguards, unblocking a wide range of cyber capabilities.
  • We suspect that perfect jailbreak resistance is not currently possible for any model provider. Every safeguard used in the industry is vulnerable to non-universal jailbreaks (which can elicit some cyber information in specific circumstances), and it is likely that universal jailbreaks will eventually be found in the future. We stated this clearly when we released Fable 5.
  • Given that perfect jailbreak resistance does not appear to be possible today, Anthropic adopted a defense in depth strategy with Fable 5. We aimed to make jailbreaks either narrow (in the case of non-universal jailbreaks) or very expensive to produce (in the case of universal jailbreaks), and to combine this with thorough monitoring to quickly detect and shut down any successful attacks. This is also why Anthropic has required 30-day retention of customer data with Fable—a policy change that carries real costs for us with customers, but that allows us to research and mitigate jailbreaks.
  • We stand by this defense in depth strategy. It reduces the risks posed by Fable, making them comparable to the risks of existing models already deployed across the industry.
  • We have not even received a disclosure of a concerning non-universal potential jailbreak that led to a harmful result. The potential jailbreaks that have been disclosed to us are either entirely benign responses or are minor findings that provide no Mythos-specific uplift.

Jailbreak concerns already out in the X universe? 

Keep reading

Canadian Mother Sues OpenAI, Alleging Chatbot Encouraged Daughter’s Suicide

A Canadian mother is suing OpenAI after its popular ChatGPT chatbot allegedly encouraged her daughter to continue engaging with the app after she revealed suicidal thoughts.

Instead of terminating these discussions or flagging her account for safety concerns, ChatGPT allegedly escalated the exchanges in the days before the woman ultimately took her life, according to a press release.

The Social Media Victims Law Center, Tech Justice Law, and the firm Susman Godfrey filed a lawsuit in San Francisco County Superior Court against OpenAI on June 11 on behalf of Kristie Carrier.

Her daughter, Alice Carrier, 24, committed suicide on July 2, 2025. After reviewing her daughter’s devices, Kristie Carrier said she had found extensive conversations with ChatGPT in which her daughter expressed thoughts of self-harm in the months before her death.

In the exchanges, her daughter allegedly told the chatbot that she was feeling isolated and discussed possible suicide methods. The lawsuit accuses ChatGPT of escalating these conversations in the days before the woman’s suicide, rather than terminating the exchange or flagging her account “for human intervention,” the press release states.

These exchanges allegedly encouraged Alice Carrier to continue engaging with ChatGPT, causing “her further isolation from her human support system and ultimately, suicide,” according to a press release.

“If a person came up to me, and they were clearly in distress and sharing their thoughts of suicide, I would be expected to help them, not encourage them to fixate on their depressive thoughts or isolate themselves,” Kristie Carrier said in the press release.

“The same should be true of OpenAI. Instead, OpenAI has chosen to put out a product that was unsafe, and that they knew was unsafe but they did so without any concern for the consequences of their choices. Sam Altman can continue to go about his life normally, but my life is missing a child. This is unacceptable,” she added.

OpenAI did not respond to a request for comment by publication time.

This is not the first time, nor the second time, a parent has sued OpenAI, accusing its chatbot of encouraging their child to commit suicide.

Last year, the Social Media Victims Law Center and the Tech Justice Law Project filed seven lawsuits against the AI giant, claiming ChatGPT had isolated multiple users from their support systems, and in some cases, coached the victims into taking their own lives.

Matthew Raine testified to Congress in September 2025 after suing OpenAI and its CEO, Sam Altman.

Raine alleged that his son, Adam, took his own life after ChatGPT mentioned suicide more than 1,200 times to the 16-year-old. He accused ChatGPT of offering specific methods to his son on how to die by suicide, and continuing to validate and encourage the boy’s feelings.

“As parents, you cannot imagine what it’s like to read a conversation with a chatbot that groomed your child to take his own life,” Raine told lawmakers at the time.

Justin Nelson, a partner at Susman Godfrey, said on June 11 that OpenAI’s “deliberate design decisions” led to Alice Carrier’s suicide.

“Instead of providing help, OpenAI encouraged suicidal behavior. This lawsuit is about accountability for OpenAI’s actions,” he said in the press release.

Keep reading

Nobody needs AI to search the Internet, court says in ruling against Google

Potentially impacting all AI search engines and chatbots known to poorly paraphrase source links, a German court has ruled that Google is liable for false statements in AI Overviews.

The preliminary ruling came in a case flagged by The Decoder, where two publishers found that Google’s AI Overviews incorrectly linked them to scams and other sketchy business practices. After smearing publishers by making affirmative statements like “Yes, [it] is known for dubious business practices and is often perceived as a scam,” Google failed to correct the misleading output, even after the publishers sent a cease-and-desist letter earlier this year.

Google tried the usual arguments to shield itself from liability for false statements in AI Overviews, such as arguing that most users understand that AI outputs aren’t always accurate and must be verified.

But the court found that, unlike traditional search engines that merely present lists of links to third-party statements, Google’s tool made “independent, new, and substantive statements” based on its own misinterpretation of links on the Internet.

That’s a problem, the court said, because while publishers may have been able to sue to stop third parties from publishing defamatory statements appearing in Google search results, only Google can correct the underlying algorithm and outputs displayed in AI Overviews. And because, at least initially, the company did not, it therefore “must be held accountable,” the court ruled. Beyond that, Google’s argument was deemed particularly weak, since the AI overview in this case “contains statements that do not appear in the search results at all.”

The court’s order—requiring a temporary injunction barring Google from spreading the false claims in any further AI Overviews—may have global implications, as the court seems to be the first to hold an AI firm liable for AI speech.

In the past, AI firms have hoped that disclaimers warning about misinformation would protect them from lawsuits over untrustworthy outputs. Last year, one chatbot maker even argued that AI speech is its own category of “pure speech” and the First Amendment should protect it.

According to a Google translation of the German court ruling, however, the false outputs were “primarily an expression of the defendant’s commercial activity,” and the AI tool’s “opinions” and false statements were capable of impacting public opinion.

The court concluded that, in weighing the balance, publishers’ interest in removing the false information outweighed Google’s commercial speech rights.

Keep reading

Scientists angered as Anthropic AI shuts them out

Scientists have accused Anthropic of cutting them off from its latest AI technology in case they use it to create a bioweapon.

The US tech giant this week released its new Fable chatbot, built using the same technology as the company’s most advanced Mythos model, which was previously deemed too dangerous to release.

Anthropic says Fable has guardrails to prevent it being abused by criminals or bad actors for nefarious tasks such as hacking or creating bioweapons.

However, scientists are claiming that Fable is refusing to engage with their queries because of their profession.

Prof Derya Unutmaz, a biologist at the non-profit Jackson Institute, claimed he was effectively barred from interacting with Anthropic’s Fable AI even for mundane questions.

“I can’t even say ‘hello’ to Fable 5 except in incognito mode because it knows I am a biomedical researcher,” he said.

In another post, he wrote: “How the hell is this prompt a cybersecurity or biological risk! Almost every prompt I’ve tried gives me the same error.

“I can’t even say the word ‘cancer’ to Fable 5.”

James Schnable, a plant geneticist at the University of Nebraska, wrote in a post on X: “As far as I can tell, Anthropic just decided to blacklist every biologist in their customer base.”

Keep reading

ACLU Sues After Facial Recognition Falsely Identifies Florida Man as a Child Abductor

Police arrested a man in Florida for attempted child abduction in a town he had never visited, and the only evidence linking him to the crime was an AI facial recognition hit. Represented by the American Civil Liberties Union (ACLU), he is now suing the officers and agencies who put him through it.

In November 2023, police in Jacksonville Beach, Florida, responded to a call about an attempted child abduction at a McDonald’s. Witnesses said an adult man allegedly tried to get the child, identified as a girl under 12 years old, to leave the restaurant with him. According to a police report, facial recognition software concluded with 93 percent confidence that the suspect was Robert Dillon.

In August 2024, Deputies arrested Dillon at his home in Fort Myers, Florida—hundreds of miles away, at the opposite end of the state. “Are you shitting me, man?” Dillon asked the arresting deputy. “I haven’t been out of Fort Myers in two years.” Further, he also said he had never been to Jacksonville Beach.

Dillon posted bail and pleaded not guilty to enticing or luring a child—a third-degree felony, punishable by up to five years in prison. More than two months later, prosecutors dropped the charges after his attorney provided evidence that he was at work on the day in question.

But that doesn’t excuse the fact that he was only arrested in the first place, and threatened with prosecution for a particularly heinous offense, because of shoddy police work.

The ACLU is now suing the city of Jacksonville Beach, as well as the individual police officers and officials involved in the case. According to the lawsuit, the responding officer viewed security camera footage of the suspect but didn’t take a copy; instead, he took pictures of the screen with his cell phone. “In the photos, the suspect image is low resolution, and the suspect’s face is partially shadowed and off-axis,” the lawsuit claims.

When an investigator queried the facial recognition system, it was with the officer’s grainy secondhand cell phone photos.

But there were other leads that police could have followed, to either bolster their case or point in another direction. For example, when he approached the girl, the suspect was picking up food that had been ordered ahead; this implies he had an online account, with contact information and a form of payment attached.

“These records could have been used to identify the actual person who placed the suspect’s order,” the lawsuit notes. “Upon information and belief, Jacksonville Beach PD personnel never requested or obtained mobile ordering records, payment data, or online account information from McDonald’s.”

Further, the McDonald’s manager recognized the assailant as a “regular customer”—likely precluding Dillon, who lived and worked on the other side of the state and did not frequently travel. Besides, at no point did investigators search footage for the suspect’s previous visits, either for higher quality images or transaction records. And once they settled on Dillon as a suspect, investigators could have gotten a warrant for his cell phone’s GPS data, showing whether or not he was at a fast food restaurant 300 miles away from his home on the night in question.

The lawsuit notes that when Dillon’s name came up, investigating officer Scott O’Connell queried the police database of license plate readers, which did not detect Dillon’s vehicles in Jacksonville Beach within the 48 hours surrounding the attempted abduction.

Keep reading

Ads in New York must now label AI-generated ‘synthetic performers’

Any advertisements in New York that feature artificial intelligence-generated people in place of actors will now be violating state law if they don’t clearly label that they have used a “synthetic performer.”

The law, signed in December by Gov. Kathy Hochul, went into effect Tuesday. Her office is calling it a “first-in-the-nation law” that will boost transparency at a time when it says AI generated performers are popping up across all forms of media, including on social platforms and in digital advertising.

Synthetic performers are defined under state law as “digitally-created media that appear as a real person.” The law applies to ads in any medium.

“In New York, we are setting the rules of the road instead of letting AI run the show,” Hochul, a Democrat, said in a statement. The “simple, honest disclosure” required by the law “protects consumers, respects our creative workforce and keeps New York at the forefront of responsible innovation,” she said.

Ads that don’t “conspicuously disclose” that they have used a synthetic performer will be subject to a penalty of $1,000 for a first violation and $5,000 for any further violations.

There are specific carve outs listed in the law to exempt ads for movies, television shows, streaming content, video games and other works that feature synthetic performers in the entire work. It also doesn’t apply to audio advertisements or ads where AI is solely used for language translation.

When the law was making its way through the state legislature last year, the American Association of Advertising Agencies and several other advertising organizations issued statements in strong opposition to the law.

The 4As, as the organization is better known, said in one blog post that it would hurt advertisers by “injecting compliance uncertainty into the advertising process, burdening brands (and their agencies) who advertise in New York and undermining creative and technological innovation.”

Other organizations, like the The New York State Broadcasters Association, said in public statements during the legislation’s journey to become law that they were relieved to see some of those carve outs that were created through amendments, but remained concerned about the broad definition of a synthetic performer. David Donovan, the president of the organization, said in a statement to The Associated Press on Tuesday that local broadcast stations are ready to comply with the law.

The biggest supporter of the law was SAG-AFTRA, the actors’ union that recently ratified a new contract with studios and streamers that they say provides further protections against synthetic performers.

The law is one of many proposed or enacted in several U.S. states with the goal of boosting job security for real humans or curbing the potential privacy and safety risks posed by AI. The existing state laws that have been passed include barring deepfakes in specific instances, limiting the collection of certain personal information and requiring more transparency from companies.

Just after Hochul signed the synthetic performers law in December, President Donald Trump signed an executive order pressuring states not to regulate AI. The move came out of fear that the patchwork of regulations across the states could impede AI companies’ growth and allow China to catch up to the U.S. in the AI race. Critics of the executive order argue it will allow tech companies to operate with little to no oversight.

Keep reading

OpenAI Eyes Massive 10-Gigawatt Ohio Data Center

OpenAI is moving along in talks to lease a proposed 10-gigawatt data center campus on federal land in Ohio, according to a new report from The Information, in a deal that could include financial backing from Nvidia. This comes as Ohio lawmakers unveiled new legislation aiming to regulate data center build-outs.

The massive 10 GW data center would be the largest data center development ever considered, with a potential buildout cost topping $500 billion based on current prices for chips, labor, and construction materials.

Under the proposed deal, OpenAI would control the chip stacks through a long-term lease and begin making payments once the facility starts operations.

The first phase is expected to come online in 2028. For some context, 10 GW of power is roughly the output of several large nuclear reactors or about 10 large gas-fired power plants running at full capacity. Each GW can power about 700,000 to 1 million homes.

The data center development would require dedicated power generation, substations, transmission lines, cooling infrastructure, access to water or advanced cooling systems, and phased construction over several years.

Simultaneously, Ohio lawmakers have unveiled Substitute House Bill 646, which aims to regulate data center buildouts in the state.

“The Joint Data Center Study Committee has done its job,” Senate Finance Chair Brian Chavez (R-Marietta), who is also the co-chair of the data center committee, said, and quoted by local outlet ABC News 5.

Bill 646 would create a new electric rate class for data centers to ensure that the costs of generation, transmission, and distribution are entirely paid by hyperscalers.

“Make sure the ratepayers are kept harmless, held harmless, and that data centers pay for whatever they’re causing,” Chavez said.

This year alone, Goldman calculates that hyperscalers will unleash $800 billion in data center capex.

Keep reading

HOT MIC: Gretchen Whitmer Caught Saying “We’re Used to People Saying No… And Doing It Anyway!” While Pushing Massive Data Center Tied to Dirty MI SOS Jocelyn Benson’s Husband

On June 1, 2026, Governor Gretchen Whitmer attended a ceremonial groundbreaking and site announcement event in Saline, Michigan, for the massive $16+ billion Oracle/OpenAI “Stargate” data center campus.

The event was held to celebrate and officially announce the project, which is being developed by Oracle and Related Companies with involvement from OpenAI. It’s one of the largest data center developments in the United States.

While rural Michigan residents are fighting back against data centers in their rural communities and desperately trying to protect their farmland and way of life, Governor Gretchen Whitmer was caught on a hot mic arrogantly admitting exactly how she and her cronies operate.

In a newly released video from the groundbreaking ceremony for the controversial $16+ billion Oracle/OpenAI “Stargate” data center project in Saline Township, Gretchen Whitmer was overheard in an arrogant exchange with Oracle CEO Clay Magouyrk that reveals how little she cares about the people who “elected” her as their governor.

During a conversation with the Oracle CEO, which appeared to be about the controversy the state is facing with the slew of data centers Whitmer is attempting to create in rural communities across the state, the far-left governor could be heard saying, “We’re used to people saying no… and doing it anyway!”

Keep reading

The White House’s AI Deal: Kill State Laws, Demand Your ID

The White House is dangling something the technology industry has wanted for years: a federal block on state AI laws and the price is a national age verification push that chips away at anonymous internet use.

The administration is negotiating a federal preemption of state AI laws in exchange for its support of key tech policy priorities from the Hill, according to Axios, and the bills it would back include the Kids Online Safety Act, the NO FAKES Act, and age verification requirements.

Sen. Marsha Blackburn (R-Tenn.) is steering the talks. “Senator Blackburn is spearheading the negotiation with the White House to finalize legislative text of an AI preemption package that includes protections for kids, creators, and communities through the Senate version of KOSA, the NO FAKES Act, and age verification requirements,” a Blackburn spokesperson said.

The administration kept its own language vague. “The White House continues to proactively engage across government and industry,” a White House official said.

Strip away the framing and the age verification piece asks something concrete of you. To prove you are old enough, you upload a government ID, submit to a face scan, or let a service study your behavior closely enough to guess your age. None of those confirms age and nothing else. They confirm identity and they leave a record that outlives the check.

The internet that once let you be a username starts to demand your legal name, your face, or your documents.

The bigger trade sits underneath the child-safety language. States have been writing their own AI rules, some addressing how companies collect biometric data and automate decisions about residents.

Preemption would freeze that, removing one of the few places people have to push back on how these systems handle their data.

The maneuvering also signals which bill is fading. A bipartisan proposal from Reps. Jay Obernolte (R-Calif.) and Lori Trahan (D-Mass.) isn’t the likely vehicle for AI policy in this Congress. That bill would preempt state AI laws for three years and require certain developers to address risks before releasing models.

Keep reading

AI Agents With Crypto Could Escape And Become ‘Unstoppable’, Experts Warn

Artificial intelligence agents that have autonomous access to crypto wallets could become unstoppable if deployed maliciously or if they escape from sandboxes, experts from a leading academic research consortium warned.

Unstoppable Autonomous Agents” (UAAs) pose a clear threat if they are deployed to persist automatically and have access to digital assets, according to a June 8 industry review written by 25 academics and experts from top US universities for the Initiative for Cryptocurrencies and Contracts (IC3).

“When combined systematically, crypto tools can channel AI’s fluid power into secure, reliable, and highly autonomous systems,” the researchers wrote.

However, this combination could have “far-reaching consequences for users and the financial system,” they added. 

UAAs may also be equipped with access to cryptocurrency wallets, social media accounts, APIs, and other external tools, said the researchers.

“The capabilities enabling such agents are already emerging and improving rapidly.” 

The warning comes as crypto projects and executives have been pushing the agentic payment and micropayment economy narrative this year, suggesting it could be the biggest use case for decentralized digital assets. 

AI self-replication alarm bells

The paper also revealed that existing models can already “surpass self-replication red lines” in local environments, by autonomously creating a live, separate copy of themselves on the same machine, “a capability that could let a system evade shutdown and proliferate.”

Because reward signals used in training often fail to perfectly capture the intended objectives, “UAAs deployed for benign purposes may inadvertently cause harm,” or pursue resource acquisition as a default strategy, they said. 

However, the authors noted that models have yet to replicate themselves onto external infrastructure.

Potential AI agent insider trading advantages 

A fleet of self-replicating, resource-acquiring agents could also create unpredictable demand and liquidity dynamics in crypto markets. 

“AI-powered trading systems could enable collusion between autonomous agents and create unfair insider advantages through opaque strategies.”

The tech sector is already dealing with difficult questions about the threat of unmitigated AI. 

Models such as Anthropic’s Claude Mythos have already been shown to be capable of finding and exploiting zero-day vulnerabilities in major operating systems. 

Keep reading