OpenAI restricts GPT-5.6 rollout following government pressure

The new lineup—comprising the flagship Sol, the balanced Terra, and the efficient Luna—was intended for a wider launch before the White House intervened. OpenAI confirmed that current access is restricted to partners whose participation has been explicitly cleared by federal authorities. This intervention follows a similar mandate against Anthropic, which was forced to pull its Fable 5 model entirely after the government restricted access for foreign nationals.

Dean Ball, a former White House AI advisor, suggests that recent executive orders have effectively institutionalized involuntary licensing. Without established safety standards, these launch delays threaten to jeopardize multibillion-dollar infrastructure investments while potentially hindering the U.S. in the global AI race. OpenAI publicly pushed back against the arrangement, stating that such government-controlled access should not become a long-term default, as it deprives developers and cyber defenders of critical tools.

To navigate these regulatory hurdles, OpenAI has integrated safety guardrails directly into the core behavior of the models. The flagship Sol model is specifically hardened against adversarial attacks, prioritizing defensive cybersecurity applications over offensive capabilities. By embedding these protections, the company aims to avoid the performance issues and user backlash that plagued Anthropic’s earlier, filter-heavy approach. OpenAI intends to move toward broader availability in the coming weeks as it negotiates a more predictable release framework with the administration.