OpenAI has released its latest advanced AI model, ChatGPT 5.6, after initially delaying the public rollout at the request of the Trump administration, which cited cybersecurity concerns. Last month the White House asked the company to restrict access to a small group of government-approved users, and OpenAI complied, briefing officials and limiting the model to trusted partners before a wider release was permitted. The episode highlights growing government intervention in the AI sector over security risks.

The staggered launch mirrors similar restrictions placed on rival firm Anthropic, whose latest models faced a temporary US export ban. ChatGPT 5.6, which includes a flagship product called Sol, is billed as OpenAI's safest and most capable model and competes with Anthropic's Claude Fable and Mythos. The wider release followed additional testing by the government's Center for AI Standards and Innovation. The restrictions have unsettled businesses, allies who felt snubbed by export limits, and observers questioning the government's regulatory approach, all as both firms jostle for position ahead of possible IPOs.