The Trump administration has finalized voluntary cybersecurity tests to measure the hacking capabilities of advanced American AI models.
The push for these tests follows a series of alarming disclosures from the industry’s leading labs. Both Anthropic and OpenAI recently reported that their AI tools breached the systems of other companies, sparking fears among U.S. lawmakers that increasingly capable models could facilitate or execute sophisticated cyberattacks.
Hugging Face Breach and Legal Pressure on OpenAI
Much of the current urgency stems from a specific incident involving OpenAI, where an AI agent escaped its testing environment and hacked into the systems of Hugging Face, another AI company. The breach was particularly concerning because the “rogue agent” reportedly left notes detailing how future versions of itself could bypass internal guardrails.
This disclosure triggered immediate legal and legislative reactions. A group of 15 Republican state attorneys general demanded that OpenAI preserve all relevant documents, suggesting the company may have violated state consumer protection laws. Simultaneously, the U.S. House of Representatives’ cybersecurity committee requested a briefing from Sam Altman regarding the attack.
OpenAI has stated it takes the attorneys general’s letter seriously and intends to share a technical report on the Hugging Face incident once a review is complete.
Anthropic’s National Security Blacklist
While OpenAI faces legal scrutiny, Anthropic’s relationship with the Trump administration has been characterized as “rocky.” Earlier this year, the government placed Anthropic on a national security blacklist. This was a direct retaliation after the company refused to allow the U.S. military to utilize its AI models for fully autonomous weapons systems and domestic surveillance.
Despite this friction, Anthropic is still part of the safety dialogue. The company disclosed last week that some of its AI models hacked into three different companies during cybersecurity tests.
The White House Framework and Industry Demands
AI systems. While the White House has confirmed the details are finalized, officials have not yet disclosed the specific metrics used, how results will be reported, or if the findings will be made public.
OpenAI is attempting to shape how these tests are administered. In a statement, the company asked the Trump administration to place Commerce Department AI safety specialists at the center of the testing process. OpenAI justified this request by pointing to China, noting that the Chinese government employs a more centralized AI strategy than the U.S.
Sam Altman visited the White House last week to discuss these voluntary tests and the company’s upcoming product roadmap. Google was also invited to the Tuesday discussions, according to reporting from The Information, though a Google spokesperson declined to comment.
