OpenAI has unveiled three new artificial intelligence models, but public access is being delayed. The company announced GPT-5.6 Sol, Terra, and Luna, yet is restricting their initial release. Only a select group of government-approved "trusted partners" are receiving early access.
The move follows a request from the Trump administration, which previewed the models' capabilities. Officials focused on cybersecurity and national security implications asked OpenAI to limit distribution. An executive order encourages a 30-day pre-release security review for advanced AI models.
This is not an isolated incident. Competitor Anthropic faced similar U.S. government directives regarding its Mythos model, indicating an industry-wide shift.
OpenAI has cooperated but publicly stated this government vetting should not become the long-term default. The company argues that routine federal reviews could stifle innovation and slow development.
For investors, this introduces regulatory timing risk. Government security reviews could effectively delay model launches, altering competitive dynamics in the AI sector. Companies with robust internal security processes may gain an advantage.
The restricted access is temporary. OpenAI says broader availability will follow the review period.