OpenAI is releasing its GPT-5.6 model, but will limit access to all three versions due to US government regulations. The move follows similar restrictions on Anthropic’s powerful AI models.
The company has released Sol, Terra, and Luna, the most powerful, efficient, and affordable versions of the new model. OpenAI aims for a broad release in the coming weeks but first wants to expand access to around 20 approved companies. The government has expressed support for the plans, but with severe restrictions that may require staggering the release.
OpenAI’s CEO Sam Altman met with White House officials earlier this month and previewed the model with the government. However, the company disagrees with the current approach and believes it is unsustainable in the long term.
The US government aims to establish a classified process by August to assess AI models’ cyber capabilities. OpenAI claims that its GPT-5.6 Sol model is better at finding vulnerabilities than carrying out attacks. The restrictions are intended to prevent prohibited offensive use, allowing for “substantial benefit” for legitimate defensive work.
Source: https://www.axios.com/2026/06/26/openai-gpt-sol-terra-luna-trump