OpenAI Limits GPT-5.6 Rollout After U.S. Government Request For Trusted Partners

OpenAI launched GPT-5.6, a new family of large language models led by its flagship Sol model, alongside Terra and Luna variants, but limited the initial rollout to a small group of trusted U.S.-based partners after a request from the U.S. government.

“Agentic AI,Artificial Intelligence & Machine Learning,Governance & Risk Management OpenAI limited its release of GPT-5”

BankInfoSecurity

The company said the GPT-5.6 series introduces a new naming system with Sol as the highest capability tier, Terra offering GPT-5.5-level performance at half the cost, and Luna targeting lower-cost, faster AI applications.

BankInfoSecurity

OpenAI said the models will become generally available through ChatGPT, Codex, and its API in the coming weeks, while also describing Sol as delivering its strongest performance yet in coding, biology, and cybersecurity.

OpenAI said it dedicated more than 700,000 A100-equivalent GPU hours to automated red-teaming to uncover jailbreak techniques before release, and it said GPT-5.6 Sol does not cross the Cyber Critical threshold under its Preparedness Framework.

In its blog post, OpenAI framed the staged approach as a short-term step, writing, "At their request, we are starting with a limited preview for a small group of trusted partners" before releasing more broadly.

Multiple outlets tied the restriction to the Trump administration’s push for pre-release review, with Reuters reporting that OpenAI restricted access to a small set of vetted partners and shared partner details with federal agencies.

Axios and The Information reported that CEO Sam Altman told employees the administration would approve customer access "customer by customer" during the preview period, and The Decoder described the same approval approach as requiring U.S. government sign-off on a case-by-case basis.

bloomingbit

OpenAI said it does not want government previews to become standard practice, writing, "We don’t believe this kind of government access process should become the long-term default."

The TechCrunch report quoted OpenAI’s Friday blog post again, saying, "We don’t believe this kind of government access process should become the long-term default," while also describing the preview as a "short-term step" toward broader availability in the coming weeks.

WIRED reported that OpenAI confirmed it would first share GPT-5.6 with a small set of customers preapproved by the US government, then work with the administration to slowly expand access.

OpenAI said GPT-5.6 Sol introduces a new maximum reasoning mode that gives the model more time to solve complex tasks, and it also launched an Ultra mode that uses subagents to tackle sophisticated workflows beyond the capabilities of a single AI agent.

“OpenAI on Friday announced three new artificial intelligence models and said it's complying with the U”

CNBC

On benchmarks, OpenAI said GPT-5.6 Sol achieved a new state of the art on TerminalBench 2.1, outperformed GPT-5.5 on GeneBench v1 while using fewer output tokens, and matched the performance of Anthropic’s Mythos Preview on ExploitBench while using roughly one-third of the output tokens.

For cybersecurity risk framing, OpenAI said, "GPT-5.6 Sol is better at helping people find and fix vulnerabilities than reliably carrying out end-to-end attacks," and it said the model still does not cross the Cyber Critical threshold under its Preparedness Framework.

The Guardian reported that European fashion retailers faced fresh questions over supply chain oversight after a fire at a factory that supplied them killed at least 33 garment workers in Bangladesh, but the GPT-5.6 rollout story emphasized how OpenAI’s staged access is meant to support broader availability while working with the administration on a repeatable process.

OpenAI also described a layered safety system combining model-level protections, real-time misuse detection, account-level monitoring, differentiated access, and extensive automated and human red-teaming, while it said it is taking the temporary step as part of ongoing engagement with the U.S. government.