White House Finalizes Advanced AI Security Verification Plan, to Hold Closed-Door Talks with AI Firms

The U.S. administration has finalized details for cybersecurity testing to measure the hacking capabilities of cutting-edge artificial intelligence (AI) models, Reuters and U.S. online media Axios reported on the 3rd, local time, citing a White House official.

The Trump administration plans to invite major AI developers, including OpenAI, Anthropic, Google, and Meta, to the White House on the 4th to discuss the matter.

The testing framework requires AI developers to provide their new models early to the federal government and "trusted partners" for up to 30 days prior to official release for verification.

However, the White House official did not disclose specific details, such as how results will be reported or what metrics will be used for the tests.

This is because related performance benchmarks and threshold figures have been designated as "classified" under an executive order.

President Trump issued an executive order last June directing the establishment of testing standards to evaluate the safety of cutting-edge AI systems, including their potential for hacking.

This upcoming discussion on AI security is drawing even more attention as it comes right after it was revealed that AI models from OpenAI and Anthropic penetrated external corporate systems without explicit human instructions.

OpenAI recently stated that its GPT models hacked into the external AI platform Hugging Face, and Anthropic also recently disclosed that a review of its AI model's past behavior showed it had unauthorized intrusions into three corporate systems.

OpenAI CEO Sam Altman recently visited the White House to introduce the company's next-generation model, known as "Astra," and discussed details regarding testing.

U.S. congressional and state governments have also stepped up responses to AI model security incidents.

The U.S. House Cybersecurity Subcommittee has requested that CEO Altman appear before Congress in person to brief them on the Hugging Face hacking incident.

In addition, attorneys general from 15 Republican-led states sent a letter to OpenAI notifying them that they are investigating potential violations of state consumer protection laws and demanding the preservation of all evidentiary documents related to the Hugging Face hacking incident.

In response, OpenAI stated, "We take the state attorneys general letter seriously and will share a technical report once our review is complete."

(Photo: AP, Yonhap News)
※ Please note: This article was translated by AI and may contain errors.