The White House Urges Anthropic to Prevent All Jailbreaks, but Success Could Be Challenging.

The Trump administration’s conflict with Anthropic regarding its cutting-edge AI models seems to be approaching a critical point.

Trump officials have informed Inner Loop that if Anthropic wishes to reintroduce Claude Fable 5—the AI model taken offline last week due to export restrictions related to jailbreaking concerns—the company must take measures to address the vulnerabilities the government claims exist.

Anthropic has insisted for several days that the administration’s worries are exaggerated and that the impact of jailbreaking is minor. The company reaffirmed this viewpoint during a technical meeting with the Commerce Department and Sean Cairncross, the Office of the National Cyber Director, on Monday.

However, officials assert that the discussion is no longer about the significance of the jailbreaks. The National Security Agency has determined that there are methods to disable the safeguards on Fable 5, which are intended to prevent access to the Mythos model’s capabilities related to cybersecurity, chemistry, and biology.

At this point, the administration essentially considers it Anthropic’s responsibility to find a solution, according to three sources familiar with the deliberations.

Neither the Commerce Department’s Center for AI Standards and Innovation nor the National Security Agency has the personnel or resources to pursue every potential jailbreak for every market model, the sources indicated.

Consequently, the administration believes that Anthropic should proactively conduct tests not only on Fable 5 but also on all of its advanced AI models to identify potential jailbreaks and report them to the government.

Yet, fundamentally, it remains unclear how Anthropic is expected to prevent jailbreaking.

Independent cybersecurity experts increasingly argue that AI model safeguards are merely temporary measures, as skilled users and emerging AI technologies will inevitably find ways to circumvent limitations—suggesting that the White House’s expectations may be unrealistic.

A spokesperson for the White House declined to make a comment.

DNI = Do Not Invite

At the beginning of the week, Trump’s selection for Acting Director of National Intelligence, Bill Pulte, appeared poised to never assume the position. Now, Trump has provided him a lifeline—and it’s the permanent DNI nominee, Jay Clayton, who now faces the possibility of not serving in the role.

To summarize: Trump initially named Pulte, his housing finance chief, to succeed outgoing DNI Tulsi Gabbard.

Confronted with bipartisan opposition due to Pulte’s lack of national security experience required by law and his controversial allegations against Trump’s political opponents, Trump nominated Clayton, the U.S. attorney for the Southern District of New York, for the permanent DNI position.

Gabbard was set to leave on June 18, with Pulte’s first day scheduled for June 19. However, Senate Republicans questioned whether if Clayton could have his hearing expedited to June 17 and begin by June 22, Pulte would even attain the position.

On Wednesday, Trump disrupted the plan. Amid a broader clash with Senate Republican leadership regarding the filibuster, Trump announced that Clayton’s hearing would be postponed indefinitely, seemingly to obstruct Pulte’s swift advancement. Senate Republicans then stated that the hearing would still occur unless Clayton failed to show up or withdrew his nomination.

The situation could be detrimental for the Office of the Director of National Intelligence, which Trump has directed Pulte to significantly downsize, and staffers have expressed dissatisfaction with what they perceive as Pulte’s lack of initiative to engage with the agency and insufficient regular briefings, according to sources familiar with the situation.