The Trump administration is limiting early access to OpenAI’s latest flagship model, GPT-5.6 Sol, as federal officials evaluate potential cybersecurity risks. OpenAI said the system will debut in a restricted preview for “trusted partners” pre-cleared with the government, with broader availability to follow. The move extends Washington’s more assertive posture toward frontier AI, following a June executive order establishing a voluntary security framework and coordination with private firms. Supporters say early vetting can mitigate national-security risks posed by models that can find software vulnerabilities; critics warn the approach could let the government effectively pick early winners in a fast-moving market. A recent episode with Anthropic—where export limits on its Fable 5 and Mythos 5 models were lifted after the company agreed to stronger safeguards—highlights the tug-of-war over access and oversight. The outcome will shape how quickly powerful AI reaches sectors such as healthcare and finance, and who gets to use it first. OpenAI has not set a general-availability date.
Related articles:
– NIST AI Risk Management Framework 1.0
– CISA Secure by Design
– Guidelines for Secure AI System Development
