Policy · 77 days ago
The new constraint is not model readiness. It is who the government lets touch the model first, which turns frontier AI release into a vetting problem as much as a product launch problem.
OpenAI said GPT-5.6 Sol will go only to customers approved by the Trump administration. Anthropic said its strongest cybersecurity model was approved for a small release to trusted partners after earlier restrictions, and the White House said it is using a 30-day cybersecurity review framework for advanced models before public release.
For AI platform teams and procurement groups, the practical effect is slower or uneven access even when vendors are ready to ship. Early availability may now favor vetted partners, and that approval gate can shape who gets to test, deploy, and compete first.
4 sources covering this story
Risky Bulletin: White House asks OpenAI to restrict GPT 5.6
The White House asks OpenAI to keep a tight grip on ChatGPT 5.6, the US Secret Service made some appalling OpSec mistakes, AMD has reintro [Read More
OpenAI voluntarily limits new AI models at government’s request
The company said Friday it was working with the government on a more formal process for reviewing model releases.
OpenAI and Anthropic Limit New AI Models to Trump-Approved Customers During Cybersecurity Review
ChatGPT maker OpenAI said Friday it is restricting the release of its new artificial intelligence model at the request of President Donald Trump’s administration.
OpenAI Previews GPT-5.6 Sol With Restricted Access and Stronger Cyber Safeguards
OpenAI previewed GPT-5.6 Sol, Terra, and Luna to approved partners, adding stronger safeguards for cyber defense use cases.
Part of the PlainSec briefing for 2026-06-29