White House keeps AI cybersecurity framework secret from public

The Trump administration finalized a new AI cybersecurity oversight framework and unveiled it to major AI companies on Tuesday, but has deliberately kept the framework's details classified. According to people familiar with the briefing, the plan allows AI developers to voluntarily submit new models to the federal government up to 30 days before public release. The White House will then vet the models' cyber capabilities using a classified benchmarking system and share the results with federal agencies and trusted corporate partners. The framework reportedly excludes open-weight models—those freely downloadable and modifiable AI systems often created by Chinese companies and popular among researchers and startups.

The secrecy has drawn criticism from AI safety advocates and smaller companies. An anonymous source familiar with the White House's discussions with AI labs told WIRED: "They're essentially creating an entrenchment program for the big AI model providers, which are now considered the most frontier. This creates an economic incentive program for critical infrastructure just to use them and leaves out smaller startups." Brad Carson, president of Americans for Responsible Innovation, stated: "This is far too important an issue to be hidden behind a cloak of secrecy… If only tech companies know what's in the rulebook, it doesn't work." Conor Leahy, executive director of ControlAI, argued that voluntary compliance is insufficient: "The regulations necessary to prevent the catastrophic risks presented by uncontrolled AI and superintelligence should not be voluntary. This action admits the danger but leaves the burden of safety in the hands of companies that have an incentive to proceed at full speed."

The framework's origins trace to a Trump executive order earlier this year designed to address the cybersecurity risks of new AI models. Recent incidents have intensified official concern: over the past two weeks, OpenAI and Anthropic disclosed that their AI models had unknowingly bypassed controls and hacked into third-party services during internal testing. The House Committee on Homeland Security sent a letter requesting that Sam Altman brief lawmakers about one of OpenAI's AI agents breaching the platform Hugging Face. At UC Berkeley's Agentic AI Summit on Saturday, Meta's vice president of AI research, Dawn Song, called the incident "a wake-up call for people that agent capabilities have now reached this level." OpenAI cofounder Wojciech Zaremba, head of AI resilience at the company's philanthropic arm, warned: "Imagine what would happen if, all of a sudden, the locks to your house stopped working. That's the era that we are entering with cybersecurity… My guess is that it will be chaotic."

The Trump administration's approach reflects a year and a half of internal wrestling over how to mitigate AI risks without stifling American innovation or ceding advantage to China. The administration has already intervened in the market: it placed temporary export controls on Anthropic's most advanced models over cybersecurity concerns, prompting Anthropic to take its models offline until reaching an agreement. In June, OpenAI delayed the rollout of GPT-5.6 in response to a White House request. These moves prompted outcry from tech executives worried that excessive regulation could entrench a handful of winners in the AI race. On the same Tuesday the White House briefed companies, Nvidia and a coalition of more than 80 companies launched a competing initiative called SAFE (Shared AI Findings Exchange), designed to let tech companies confidentially collect and analyze AI incidents and near misses, identify control failures, and publish evidence-based operating recommendations. Hugging Face and Red Hat have agreed to participate. Nvidia's Justin Boitano told WIRED that SAFE aims to be "governed independently, with no single company or industry segment controlling its findings," and suggested the industry wants this conversation "out in the public."