AI Giants Collaborate with U.S. Government on Safety Reviews

Microsoft, Google’s DeepMind, and Elon Musk’s xAI have committed to sharing early versions of their advanced AI models with the U.S. government for initial reviews and safety evaluations, as announced by the Commerce Department on Tuesday.

The department’s AI Standards and Innovation Center will carry out “pre-implementation assessments and targeted research” aimed at gaining a better understanding of these new technologies’ capabilities and potential risks.

In addition, prior agreements with Anthropic and OpenAI have been “renegotiated” to align with new national security directives from Commerce Secretary Howard Lutnick and former President Trump.

This initiative comes in response to rising concerns about emerging AI tools like Anthropic’s Mythos, which company executives caution could lead to security threats such as hacking and terrorism if misused.

Google DeepMind has chosen not to comment, though Tom Lu, its vice president of global AI, confirmed the partnership publicly on Tuesday.

As of the announcement, responses from the White House, Commerce Department, Microsoft, and xAI are still pending.

Chris Fall, director of the Center for AI Standards and Innovation (CAISI), emphasized the importance of “independent, rigorous measurement science” for understanding the implications of cutting-edge AI on national security.

These new collaborations are expected to bolster public interest efforts at crucial times, he noted.

Interestingly, it was revealed that Fall would take over as director after Colin Burns, a former anthropologist, left the role just days into his tenure due to friction over security strategies with Anthropic.

The AI Safety Institute was formed in 2023 during the Biden administration and was previously modified under Trump’s leadership to hasten AI deployment while aiming to lower safety barriers.

Trump had voiced the importance of rapidly advancing technological capabilities to outpace China in the global AI landscape.

However, the contentious introduction of Mythos by Anthropic has raised alarms among some policymakers.

A disturbing analysis from Anthropic indicated that if Mythos were to be breached, it could potentially take control of critical infrastructure like power grids and hospitals.

The firm highlighted that the model has already identified numerous serious vulnerabilities across major operating systems and web browsers, with access restricted to a select group of companies including Amazon and JP Morgan.

Anthropic’s CEO, Dario Amodei, suggested that competitors might catch up swiftly in the coming months.

Meanwhile, OpenAI is set to release its latest model, GPT-5.5-Cyber, on a limited basis, citing security considerations.

It appears that public sentiment towards AI is growing increasingly skeptical. A recent survey conducted by the Pew Research Center revealed that about half of both Republicans and Democrats expressed more concerns than excitement regarding AI’s escalating role in daily life.