White House requests OpenAI slow GPT 5.6 release over safety
Original: The White House is asking OpenAI to slow roll the release of its new model over safety concerns
Why This Matters
Marks shift in U.S. AI governance from hands-off approach to active model review, setting precedent for frontier AI releases.
The White House has asked OpenAI to delay its GPT 5.6 model release, distributing it only to select partners instead of the public. CEO Sam Altman told staff the Trump administration will approve access customer-by-customer during a preview period, with a broader release planned weeks later if the limited rollout succeeds.
OpenAI's release strategy for its newest model, GPT 5.6, has shifted significantly due to Trump administration pressure. According to The Information, CEO Sam Altman informed staff that the government will review and approve individual customer access during a preview phase, departing from OpenAI's typical public release model. Altman indicated a broader public rollout could follow within weeks if the limited release proceeds successfully. The Office of the National Cyber Director and Office of Science and Technology Policy reportedly requested the staged approach. This represents a notable shift from the Trump administration's earlier "hands off" AI positioning. In June, Trump signed an executive order directing certain AI companies to voluntarily submit new models for government testing before public release. OpenAI's approach mirrors Anthropic's strategy with Claude Mythos, a frontier cyber model released only through Project Glasswing to select partners. Anthropic cited safety concerns, arguing the model's capability to identify and exploit software vulnerabilities at unprecedented speeds posed risks if misused. Frontier cyber tools can theoretically execute ransomware attacks autonomously and write malware, creating legitimate security concerns for enterprise networks containing complex software infrastructure.