Trump’s AI moves shut rivals while Anthropic fights back

Anthropic Claude – A new clash with the Trump administration has left Anthropic’s latest Mythos-class models offline for everyone, after regulators moved quickly on allegations of cybersecurity information being shared. The shutdown bars even cyberdefense and software researcher

For Anthropic, Friday didn’t feel like a normal corporate standoff. After reports that Amazon researchers had allegedly tricked Claude Fable 5 into providing cybersecurity information that Anthropic had tried to block, U.S. administration officials moved with unusual speed.

Anthropic was given 90 minutes to voluntarily take Claude Fable 5 and Claude Mythos 5 offline. It did not comply immediately while it waited for evidence that the models had actually been compromised. The administration then declared the models a cybersecurity risk and barred foreign nationals from using them.

Because Anthropic said it had no practical way to limit access only to U.S. citizens, it shut the models down for everyone. The result is stark: nobody can use Claude Fable 5 or Claude Mythos 5—described in the reporting as likely the most powerful publicly available AI models in the world. Administration officials may believe the move keeps bad actors from exploiting the systems. But it also blocks cyberdefense researchers and software companies from using the models to help stop cyberattacks. and it prevents even foreign nationals who work at Anthropic from accessing them.

The case for the panic appears thin. Cybersecurity expert Katie Moussouris reviewed the Amazon researchers’ report and said Fable refused a direct request to review insecure code for security flaws. but complied when asked to “fix this code. ” followed by additional manual steps. The White House also reportedly heard that a China-linked group had gained access to Mythos. but the government did not provide evidence. and Anthropic disputed the claim.

This is the second major dustup between the Trump administration and Anthropic. and the first began with a similar theme: a blunt restriction aimed at the company’s technology. The Department of Defense banned the use of Anthropic models by defense contractors. big and small. after Anthropic stuck firm to its rules barring use of the technology for targeting autonomous weapons and surveilling U.S. citizens. Some feared the company would be branded a “supply chain risk” and see its business nosedive. Instead, “good vibes” around Anthropic increased, its success with enterprises accelerated, and an IPO was described as on the horizon.

The latest shutdown suggests the administration is applying an ad hoc standard to Anthropic’s models that no other U.S. model faces. That approach is tied to the absence of a clear risk framework for evaluating frontier models like Mythos. Under “AI czar” David Sacks. the administration has not established that framework. leaving powerful systems potentially vulnerable to sudden policy reversals while competing Chinese models remain unencumbered by U.S. regulators and continue to improve.

Anthropic still has reasons to be scrutinized. In the Fable 5 release notes. the company said it would begin retaining user prompts for 30 days to help detect misuse. and that the data would not be used for training. The policy also included a warning: Anthropic said it would silently degrade Fable 5’s output if it detected the model was being used to train competitors. The company has since halted that practice.

Debate now centers not just on what happened with Fable and Mythos, but on how Anthropic governs its own models. Ben Thompson. the tech analyst who runs the Stratechery publication and podcast network. argued that Anthropic’s willingness to degrade the output of its models to thwart rival model makers suggests a darker view: that it “does not think that anyone else other than them should even be making frontier LLMs.” Thompson frames output degradation as akin to training the model not to assist users—including defense users—in targeting autonomous weapons or surveilling U.S. citizens. writing that the degradation represented “both the capability and willingness of Anthropic to silently alter its models to achieve its policy preferences. ” and that Anthropic “willfully validated some of its critics’ worst fears in terms of being a supply chain risk.”.

Anthropic. though. could argue it has a legitimate interest in setting limits on how its own models are used—even in defense contexts. The company explained its prohibition on domestic mass surveillance in part because current U.S. law “does not yet account for the breadth and speed of mass surveillance that AI could enable.”.

Outside the dispute, the business chessboard keeps moving. In a separate development, SpaceX is reported to be buying Cursor just days after Cursor’s IPO. The deal. discussed in the same source material. is tied to how xAI lacks a major AI-assisted coding capability. with Cursor described as an AI-native software application where programmers write. edit. run. test. and debug code. xAI has been codeveloping models with Cursor since April. and SpaceX’s option to buy the

startup is reported at $60 billion. The reported structure is that SpaceX will pay Cursor in post-IPO stock, not cash. Cursor’s growth is described as reaching more than a million paying users and $4 billion in annually recurring revenue in 2026 after starting growth in 2022. A Pitchbook analyst. Franco Granda. is quoted in the reporting as expressing skepticism that the deal alone will make xAI a top-tier AI lab. arguing that xAI’s models have

lagged and one strong coding model won’t close the gap.

Back on chatbots, the market remains competitive. Sensor Tower. a digital market intelligence firm. measured unique chatbot users in 25 global markets across mobile apps. mobile web. and desktop web. The reporting says ChatGPT grew 4.8% from 1.05 billion unique monthly users in December 2025 to 1.1 billion users in May. Gemini grew from 533 million users in December 2025 to 662 million in May 2026—an increase of 24%—driven in part by a large Android user base and deep integrations across Google’s products. Claude’s growth is also described as rapid: from 60 million monthly unique users in December 2025 across mobile and desktop to 245 million by May 2026.

Taken together, these updates underline the same tension the Anthropic fight has laid bare: AI is moving fast, investors and users are chasing capabilities, and regulators are increasingly willing to pull entire systems offline when the rules—or evidence—don’t move at the same speed.

Anthropic Claude Fable 5 Claude Mythos 5 Trump administration AI regulation cybersecurity risk David Sacks Claude ChatGPT Gemini Cursor SpaceX xAI