OpenAI Limits GPT-5.6 Rollout After Government Request

OpenAI has partially restricted the rollout of its latest model, GPT-5.6, following a request from the U.S. government. While the company complied with the request, its leadership has publicly pushed back, arguing that such government intervention should not become a recurring pattern in AI development. The specific restrictions affect certain advanced capabilities of GPT-5.6, particularly those related to automated code generation for critical infrastructure and real-time data analysis at scale. OpenAI did not disclose the exact nature of the government’s concerns but stated that the limitations are temporary while both parties work on a framework for responsible deployment. In a blog post addressing the situation, OpenAI emphasized that government access processes—where agencies can request limitations on AI capabilities—risk keeping the most powerful tools away from the users and developers who need them most. The company argued that such restrictions could slow innovation in fields like medical research, climate modeling, and cybersecurity. This is not the first time a major AI company has faced government pressure. Earlier this year, similar requests were made regarding models capable of generating synthetic media or automating certain cybersecurity tasks. However, the GPT-5.6 case is notable because the model had already been tested by thousands of developers before the restrictions were applied. Industry reactions have been mixed. Some AI safety advocates applaud the government’s caution, pointing to potential risks of unconstrained AI capabilities. Others worry that piecemeal restrictions could create an uneven playing field, where U.S. companies face limitations while international competitors operate without similar constraints. OpenAI has proposed a more structured approach: instead of ad-hoc government requests, the company wants clear, publicly available guidelines for what constitutes a national security concern in AI. The company also suggested that any restrictions should come with sunset clauses and regular reviews. For now, GPT-5.6 remains available with the restricted features disabled. Developers who had already integrated those features will need to adapt their applications. OpenAI has promised to release a detailed technical report on the model’s capabilities and the nature of the restrictions in the coming weeks.