White House asks OpenAI to slow GPT-5.6

This is the first time federal units have directly intervened to slow a frontier model's rollout, signaling a new era of hands-on government oversight for AI deployment.

• –The intervention was led by the Office of the National Cyber Director and the Office of Science and Technology Policy under a new executive order.
• –Shifting to a customer-by-customer approved preview allows OpenAI to comply with federal pressure while keeping a foot in the door for commercial testing.
• –The move follows a recent global shutdown directive for Anthropic's Fable 5 and Mythos 5 models, indicating a broader, regulatory crackdown on unaligned agentic capabilities.
• –For developers, this means access to next-generation frontier models will likely be gated, slow, and heavily audited for safety and exploit potential.