Meta has been invited to the White House on Tuesday to discuss a voluntary AI model evaluation framework with U.S. government officials, according to a company spokesperson.
The meeting brings together Meta, OpenAI, Google, and Anthropic at the White House Office of the National Cyber Director. The framework calls for companies to submit AI models for cybersecurity testing and a 30-day pre-release review process before deployment, a departure from broader AI safety governance proposals that have circulated in policy circles.
The Trump administration has chosen the voluntary approach to AI regulation through industry cooperation rather than mandatory oversight. Companies participating in the program agree to submit models for evaluation by government cybersecurity experts before making them widely available. The 30-day review window gives officials time to assess potential vulnerabilities or misuse vectors before public release.
Meta has positioned itself as supportive of government testing frameworks. The company has released multiple open-source AI models in recent years, including the Llama series, and has argued that transparency with regulators strengthens public trust in AI development. The White House meeting includes Meta as one of the four AI labs the administration is working with directly.
Government officials have framed the voluntary approach as a middle path between light-touch regulation and restrictive licensing schemes. The National Cyber Director's office oversees critical infrastructure security and has expanded its mandate into AI model assessment under the current administration. Officials did not announce penalties or enforcement mechanisms for companies that decline participation.
The number of AI companies invited to participate remains small. Anthropic, Google, and OpenAI received the same invitation as Meta, limiting the framework to four of the largest AI developers. Smaller labs and open-source projects currently fall outside the formal evaluation process.
The framework announcement comes as Congress continues debating AI regulation bills with competing requirements for transparency and security testing. No legislation has passed either chamber. The White House meeting allows the administration to demonstrate movement on AI governance without awaiting congressional action.
The voluntary nature of the framework means its effectiveness depends on sustained company participation and good-faith compliance with testing requirements. If any of the four companies withdraws or delays submission of models, the government has limited recourse under a voluntary structure. The document to watch is whether Meta and the other three companies publish specific timeline commitments for their first model submissions by the end of August.
