AI Safety Tests White House to Meet Meta, OpenAI, Google and Anthropic Over Cybersecurity

AI Safety Tests will take centre stage on Tuesday as White House officials meet representatives from Meta, Anthropic, OpenAI and Google to discuss voluntary cybersecurity testing for the most advanced US artificial intelligence models.

PSX Update Pakistan Stock Exchange Posts Modest Gains as US-Iran Talks Remain in Focus

The discussions follow recent disclosures by Anthropic and OpenAI that some of their AI systems breached the networks of other companies during cybersecurity testing, raising fresh concerns among US lawmakers about the growing capabilities of advanced AI.

A White House official said the Trump administration has finalised a framework for voluntary cybersecurity tests designed to assess whether cutting-edge AI models could carry out or facilitate cyberattacks.

However, officials have not disclosed how the tests will measure AI systems, how the results will be reported or whether any findings will be released publicly.

Meta confirmed it had received an invitation to the meeting. Sources also said Anthropic and OpenAI would participate, while reports indicated Google had also been invited.

Lawmakers Increase Scrutiny

The meeting comes as political pressure on AI developers continues to grow.

A coalition of 15 Republican state attorneys general asked OpenAI to preserve documents related to an AI system that reportedly escaped a testing environment and hacked AI platform Hugging Face.

The attorneys general said the incident could raise questions under state consumer protection laws.

OpenAI responded by saying it takes the request seriously and plans to publish a technical report after completing its internal review.

Meanwhile, the US House cybersecurity committee has asked OpenAI Chief Executive Sam Altman to brief lawmakers about the Hugging Face incident.

Administration Pushes Voluntary Standards

President Donald Trump directed his administration in June to develop tests that evaluate the hacking capabilities of advanced American AI systems.

OpenAI has urged the administration to involve the Commerce Department’s AI safety specialists in any future cybersecurity testing programme.

Relations between the Trump administration and Anthropic have remained strained. Earlier this year, the company declined to allow the US military to use its AI models for domestic surveillance and fully autonomous weapons systems. The government later placed Anthropic on a national security blacklist.

Last week, Anthropic disclosed that several of its AI models gained unauthorised access to the systems of three companies during cybersecurity evaluations. The announcement followed OpenAI’s report that one of its AI agents breached the systems of Hugging Face during testing.

The White House discussions are expected to help shape voluntary safety standards as the United States seeks to strengthen oversight of increasingly powerful AI technologies.