The Trump administration has prompted OpenAI to restrict access to its newest language model, GPT 5.6, by sharing it only with a limited group of partners during an initial preview. This cautious approach marks a shift from OpenAI's typical wider public releases.
- The Trump administration is overseeing GPT 5.6's preview access.
- OpenAI's release strategy aligns with recent federal AI oversight actions.
- Safety concerns center on powerful AI’s potential cybersecurity risks.
What happened
OpenAI has announced it will initially limit access to its new GPT 5.6 model to a small group of trusted partners rather than releasing it directly to the public. This decision follows direct requests from the Trump administration, which is now playing an active role in regulating the rollout of advanced AI models.
CEO Sam Altman disclosed that the government will vet customers on a case-by-case basis during the model’s preview period. If this controlled release goes smoothly, OpenAI plans to consider a broader public launch in the coming weeks. This marks a notable shift from previous OpenAI launches, which tended to be more open.
Why it matters
The administration’s intervention underscores growing federal concerns about the risks of powerful AI technologies. Despite initially advocating a hands-off approach to AI, recent policy moves, including an executive order mandating voluntary pre-release government testing, indicate a pivot toward stricter oversight to prevent misuse and cybersecurity threats.
Notably, AI models like GPT 5.6 and Anthropic’s Claude Mythos have raised alarms about their potential for misuse, such as facilitating sophisticated cyberattacks. These frontier models are capable of rapidly identifying and exploiting software vulnerabilities, which could pose significant threats to enterprise and government systems if widely accessible.
What to watch next
Observers will be closely monitoring how the controlled rollout proceeds and whether the government’s oversight successfully mitigates risks without stifling innovation. The outcomes of this preview phase could set important precedents for how future AI models are released and regulated in the US.
Additionally, the broader industry trend of limiting initial access to powerful AI tools, as seen with Anthropic’s Project Glasswing, may become a new norm. The balance between unlocking AI’s potential and ensuring safety will remain a key challenge for regulators, developers, and users alike.