Axios reported that the Trump administration has given OpenAI clearance for a broad launch of GPT-5.6 after additional testing and meetings with government officials. OpenAI had previously limited access to government-approved users, a release structure the company said was not its preferred path. The wider launch is now expected after review by the Commerce Department's Center for AI Standards and Innovation.
The OpenAI decision fits a broader pattern. AP reported that the administration recently lifted restrictions on Anthropic's Claude Fable 5 while keeping its more powerful Mythos 5 limited to approved U.S.-based organizations. The Anthropic restrictions had been tied to cybersecurity concerns after researchers found ways the model could help discover and potentially exploit software vulnerabilities.
This is AI governance in an interim state: powerful models are being reviewed case by case, with companies, federal agencies, and security experts negotiating access before detailed standards are complete. That may be pragmatic for fast-moving systems, but it can also create opacity, inconsistent treatment, and incentives for firms to solve policy through private meetings rather than public rulemaking.
The responsible question is not simply whether a model is released. It is what evidence justified release, who can audit that evidence, how restrictions are applied across competitors, and whether users understand the difference between safety testing, national-security screening, and market timing.