The White House has pushed back against expanding access to Mythos, citing national security and computing constraints, according to a The Wall Street Journal report. The move highlights growing tensions between innovation and safety as powerful AI tools capable of cyberattacks come under tighter scrutiny.

The race to build ever more powerful artificial intelligence is no longer just a contest between tech companies, it is fast becoming a matter of state control. At the centre of the latest flashpoint is Mythos, a highly advanced system capable of identifying and exploiting software vulnerabilities at scale.

According to a report by The Wall Street Journal, the White House has opposed a proposal by Anthropic to expand access to the model, complicating plans to put the technology in the hands of a broader group of companies and organisations.

STORY CONTINUES BELOW THIS AD

The resistance signals a deeper unease within government circles.
Mythos is not just another AI product launch, it is a tool that could fundamentally reshape cybersecurity, for better or worse. And for policymakers, that dual-use nature is proving difficult to manage.

Security fears meet political tensions

Anthropic had proposed increasing access to Mythos from roughly 50 organisations to around 120, including companies tied to critical infrastructure. But US officials pushed back, citing concerns that a wider rollout could heighten security risks.

The model’s capabilities have already unsettled agencies. By rapidly detecting and chaining together vulnerabilities,
Mythos could enable sophisticated cyberattacks at a scale that outpaces current defence mechanisms. While it can also be used to strengthen systems, the fear is that the same power could be misused if it spreads too quickly.

There are also practical constraints. Some officials are concerned that Anthropic may not yet have the computing capacity required to support a larger user base without affecting government access. In a landscape where processing power is as strategic as the software itself, prioritisation becomes a sensitive issue.

The standoff reflects a broader, and at times strained, relationship between Anthropic and the US administration. Previous disagreements, including disputes linked to defence applications of AI, have spilled into legal battles, adding a political dimension to what might otherwise be a technical debate.

Criticism of Anthropic has also extended beyond security. Some in the administration have taken issue with the company’s perceived political leanings and its links to figures associated with earlier governments, further complicating cooperation.

STORY CONTINUES BELOW THIS ADInnovation, access and the limits of control

Despite the pushback, the
White House remains closely involved in Mythos’s rollout, underscoring the model’s national security implications. A limited number of organisations, including some government agencies, already have access, and discussions are ongoing about how to expand that safely.

Anthropic, for its part, has emphasised caution. It has no immediate plans for a public release and has instead focused on controlled access through initiatives that allow select partners to test the model’s capabilities. Even so, concerns have grown after the company revealed it was investigating possible unauthorised access to the system.

Across the industry, similar tensions are emerging. Leading AI firms such as OpenAI and Google have also begun granting early access to security researchers, hoping to identify and patch vulnerabilities before they can be exploited. Yet the scale of the challenge is daunting. As AI systems uncover thousands of new bugs, the task of fixing them all, quickly and effectively, becomes increasingly complex.

Analysts warn that a fragmented approach could be counterproductive. Cooperation between governments and AI companies, they argue, is essential to managing risks that no single entity can handle alone.

At the same time, the issue of computing power looms large. Expanding access to advanced models requires vast infrastructure, something even well-funded companies are still building out. Anthropic has recently struck deals with major technology players to boost its capacity, but these efforts will take time to come online.

STORY CONTINUES BELOW THIS AD

First Published:
April 30, 2026, 13:12 IST

HomeTechWhite House pushes back on wider rollout of Anthropic’s Mythos AI over security fearsEnd of Article