The White House has launched a new initiative called "Gold Eagle," which will coordinate the handling of software vulnerabilities discovered through AI — and may also oversee which companies and institutions get early access to cutting-edge AI models.
CNBC reports, citing sources familiar with the matter, that the program could be responsible for approving the first partners for frontier AI models from companies like OpenAI and Anthropic. In the future, when these companies choose their initial users, they might need explicit government approval.
The US government has already been involved in vetting the first partners for GPT‑5.6, and such ad‑hoc arrangements could now be folded into a formal framework under Gold Eagle.
However, the White House told CNBC that it does not have approval power. Participating in testing remains voluntary, and companies still have the final say on when to release a model and who gets access.