By
OpenAI said that it plans to give a select group of government-approved partners early access to GPT-5.6 Sol, its most powerful AI model to date, before releasing the product more broadly. The limited rollout follows a request from the Trump administration, which asked the company to adopt a phased launch strategy for its next-generation AI system while security reviews are conducted.
Last week’s request came from the White House’s Office of the National Cyber Director and Office of Science and Technology Policy, which have pushed AI developers to give federal agencies early access to frontier models so officials can evaluate their capabilities and potential security risks before wider deployment,
OpenAI CEO Sam Altman told employees in a memo that GPT 5.6-Sol will initially be available to 20 approved partners, including Amazon’s Bedrock platform. According to the memo, access is being granted on a “customer-by-customer” basis while the review process is underway.
“We’ve made clear to the U.S. government that this is not our preferred long term model, and will work with them and others in industry to achieve a more sustainable approach for future releases,” Altman said in the memo. Altman said he hopes to release GPT-5.6 to the public a “couple of weeks later.”
The decision reflects growing concern over what the Trump administration says are legitimate national security implications of increasingly capable AI systems.
OpenAI says GPT-5.6 Sol is its most advanced model to date, with improvements in reasoning, autonomous task execution, software engineering and cybersecurity-related capabilities. It released benchmarks that said its performance was broadly comparable to Anthropic’s Mythos 5, which was withdrawn on June 12 following a directive from the Commerce Department expressing concerns that its advanced capabilities could create new cybersecurity threats.
At the time, Anthropic said the concerns were over “a small number of previously known, minor vulnerabilities” and that “other publicly-available models are able to discover them as well without requiring a bypass.”
Politico reported that the initial vulnerability was brought directly to the White House by Amazon CEO Andy Jassy. (Amazon is an investor in Anthropic.)
“We have reviewed a report that we believe is the basis of the government’s directive and validated that the level of capability displayed there is widely available from other models (including OpenAI’s GPT-5.5), and is used every day by the defenders who keep systems safe,” Anthropic said in its statement.
Anthropic and the Trump administration have been sparring for months since a dispute over how Anthropic’s models would be used by the Pentagon. Reports that the company’s models were used in U.S. operations in Venezuela in January, in violation of its licensing terms, which prohibit using Claude models to commit violence, led to a near complete rupture.
Adding further layers to the drama, just hours after OpenAI announced the limited rollout of GPT-5.6, Anthropic disclosed that the Trump administration had approved a limited release of Mythos 5, reversing the Commerce Department restriction.
The restrictions on GPT-5.6 Sol and Mythos 5 follow a recent call from the Five Eyes alliance for closer coordination on advanced AI development and security. A White House official said the administration continues “to collaborate with frontier AI labs to develop shared approaches for addressing the challenges of scaling this technology.”
