Phase-In Order: US Government Mandates Staged Release for OpenAI's GPT-5.6

According to a person familiar with the discussions, the US government has directed OpenAI to adopt a phased release strategy for its upcoming next-generation AI model, GPT-5.6. The instruction, originating from the Trump administration, signals a significant escalation in regulatory scrutiny over the potential impacts of cutting-edge artificial intelligence.

A Controlled Rollout Plan

OpenAI CEO Sam Altman relayed this information to employees during a recent internal meeting. As per the government's request, GPT-5.6 will initially be made available only to a select group of trusted partners for evaluation. Broader access to the model will be granted incrementally only after this preliminary phase.

Altman explained that US officials are growing "increasingly concerned" about the capabilities and potential risks associated with frontier AI models. He emphasized that while OpenAI might not agree with every regulatory viewpoint, the company is committed to cooperating fully with government stipulations regarding model safety and deployment safeguards.

Industry Context: Ripples from Anthropic's Move

The timing of this intervention is notable. It comes less than two weeks after OpenAI's rival, Anthropic, paused access to its own advanced model under regulatory pressure. Although Altman did not explicitly reference this event, industry analysts see it as a catalyst that heightened regulatory urgency for controlled deployments across the sector.

These consecutive developments point to an emerging pattern: US regulators are seeking to establish more rigorous, pre-deployment controls for AI models with breakthrough capabilities. A phased release serves as a compromise tool, aiming to strike a balance between rapid innovation and risk mitigation.

Looking Ahead: A New Equilibrium for AI Deployment

This directive could set a precedent for how powerful AI models are brought to market. Other companies developing systems with comparable capabilities may face similar pre-launch reviews or staged release requirements in the future.

  • Adjusted Development Cycles: AI firms will need to factor external safety evaluations and regulatory dialogues into their product timelines.
  • Partner Vetting: The definition and selection criteria for "trusted partners" will likely become a key point of discussion between companies and regulators.
  • Global Ripple Effects: The US approach may influence regulatory thinking on generative AI in other jurisdictions.

Neither OpenAI nor relevant US government agencies have publicly commented on the matter. However, it is evident that regulatory hands are now gripping the steering wheel of AI development more firmly, attempting to set a safe pace between breathtaking innovation and uncharted risk.