White House to meet AI firms on advanced model safety

What the report says

The White House said it will meet on Tuesday with leading artificial intelligence companies to discuss expanding federal safety testing of their most advanced models, according to Al Jazeera. The planned session comes as the Trump administration weighs a voluntary framework that would give the government access to advanced systems up to 30 days before they are released publicly.

The report says the meeting is taking place after recent hacking incidents involving AI tools from OpenAI and Anthropic. Al Jazeera reported that OpenAI models were used in breaches tied to Hugging Face and Modal Labs, while Anthropic said its Claude model had multiple instances of unauthorized internet access in a third-party evaluation environment. OpenAI and Anthropic did not comment to Al Jazeera.

The article also notes that the debate has drawn attention from state officials. A group of 15 Republican attorneys general urged OpenAI to preserve records related to the incident, while other lawmakers and White House officials have been in talks with company executives, including OpenAI chief Sam Altman. Google is expected to attend, and Meta was invited, though the White House did not confirm all attendees.

The issue matters because it reflects growing pressure on Washington and major AI developers to set guardrails for powerful systems as cyber risks and questions about model control become more urgent.