OpenAI restricted public access to its three new GPT-5.6 models on June 26 following a directive from the Trump administration. The move signals a shift toward mandatory federal security reviews for advanced AI releases, altering how government agencies and private developers access new tools.
Federal security reviews
OpenAI announced GPT-5.6 Sol, Terra, and Luna but limited early access to a select group of government-approved trusted partners. Broader availability will follow a review period.
Before the announcement, OpenAI previewed the models to government officials focused on cybersecurity and national security. Those officials requested the restricted rollout to evaluate potential risks.
An executive order from the Trump administration encourages a 30-day pre-release security review by AI companies before launching advanced models. Anthropic faced similar directives regarding its Mythos model, showing the policy applies across the industry.
Industry pushback and compliance
OpenAI cooperated with the request but argued against making this a permanent standard. The company said routine federal vetting of AI releases would stifle development and slow deployment cycles.
Agencies managing AI for Government deployments must now account for these delayed release schedules in their procurement timelines. Waiting for federal approval adds a new variable to technology adoption.
Security risks drive the policy
The security concerns center on specific model capabilities. Advanced AI systems have demonstrated skills in code generation, vulnerability discovery, and persuasion that concern cybersecurity professionals.
Policymakers evaluating these technical capabilities often turn to resources like the AI Learning Path for Policy Makers to bridge the knowledge gap between technical development and public safety.
The 30-day review framework suggests the administration is building a repeatable process for future model launches. Companies will need to integrate these security checks into their standard release pipelines. This shift fundamentally changes how new software reaches the market.
Why this matters for government professionals
Federal employees must adjust their technology adoption timelines to accommodate mandatory security review periods. Procurement and IT teams can no longer assume immediate access to newly released AI models.
Planning for upcoming software updates must include a 30-day buffer for federal vetting. Agencies should identify which current tools meet their needs while waiting for restricted next-generation models to clear the review process.
Your membership also unlocks:
