White House Finalizes Secretive AI Safety Vetting Framework
The Trump administration has finalized a framework with major technology companies to vet new artificial intelligence models for safety and cybersecurity risks. However, the White House has chosen to keep the policy private, sharing testing criteria exclusively with select firms rather than the public, raising transparency concerns among observers.
Key points
- The Trump administration finalized a voluntary framework for testing new artificial intelligence models for cybersecurity and safety risks.
- Representatives from OpenAI, Anthropic, Meta, Google, Nvidia, and Microsoft attended a private White House meeting on Tuesday.
- The White House declined to release the policy publicly, opting instead to share testing criteria only with select technology companies.
- Discussions for the framework originated earlier this year following Anthropic's decision in April to withhold its Mythos model due to hacking risks.
The White House has concluded months of discussions with artificial intelligence industry leaders by finalizing a new framework designed to test emerging AI models for safety and cybersecurity threats. Despite the finalized agreement, the administration has opted to keep the operational details private, withholding the policy from public view and limiting the distribution of testing criteria to a select group of firms.
On Tuesday, officials from major technology enterprises, including OpenAI, Anthropic, Meta, Google, Nvidia, and Microsoft, participated in a closed-door meeting with White House staff to review the established framework. The decision to maintain secrecy around the testing procedures has drawn scrutiny regarding transparency, leaving external businesses, foreign governments, and the public without insight into the specific safety benchmarks models must satisfy.
Efforts to establish a cybersecurity vetting framework for artificial intelligence began earlier this year. Those discussions were spurred by the development of Anthropic’s Mythos model, which was withheld from public release in April over concerns regarding its potential application in compromising financial and information technology systems.
Sources
The WireByte editorial team synthesises technology news from multiple primary sources, verifies the facts, and links every source. Articles are produced with AI assistance and reviewed under our editorial policy.