A government asked OpenAI to limit access to GPT-5.6, and OpenAI obliged — then published a statement signaling it doesn't want that to become a template.
OpenAI confirmed it restricted the rollout of GPT-5.6 at the request of a government, though it did not name the government or specify what the restrictions involved. The company complied but wasted little time distancing itself from the arrangement. "We don't believe this kind of government access process should become the long-term default," it said, adding that the current setup "keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them."
The episode puts OpenAI in an uncomfortable position it has occupied before: following government direction while publicly arguing against the precedent it just set. The concern carries real weight — access controls imposed on frontier models can cut off exactly the security researchers and enterprise customers who stand to gain the most from them.
By objecting on the record, OpenAI signals to other governments and to its own paying customers that this was a one-time accommodation, not a standing policy. Whether that framing survives the next request is a different question.