OpenAI finally released its advanced ChatGPT 5.6 model on Thursday, following a delay triggered by U.S. government cybersecurity concerns. The Trump administration had asked the company last month to limit the initial rollout to a small circle of government-approved users. OpenAI complied, restricting access to trusted partners while briefing officials on the model’s capabilities. The wider release came only after additional testing by the government’s Center for AI Standards and Innovation.

The staggered launch mirrors restrictions placed on rival Anthropic’s latest models, which faced a temporary export ban last month. Both incidents have spotlighted the cybersecurity risks tied to AI, tensions with U.S. allies over sharing breakthroughs, and criticism of the government’s ad-hoc regulatory approach. ChatGPT 5.6 includes a flagship product called Sol, which OpenAI touts as its safest and most capable model yet. It competes directly with Anthropic’s Claude Fable and Mythos models, which hit the public earlier this month after their own government review.

The White House has largely pushed for rapid AI development, citing the need to outpace China, but recent fears of destabilizing AI-driven cyberattacks have spurred more intervention. President Trump issued an executive order last month calling for voluntary government reviews of new models. The restrictions on access have frustrated industries that depend on these tools, as well as allies excluded from the initial rollout. Meanwhile, OpenAI and Anthropic are racing toward initial public offerings later this year—Anthropic recently hit a $965 billion valuation, while OpenAI is reportedly seeking $1 trillion, though its IPO may slip to next year amid global tech market turbulence.

Source: The Guardian