IT·Science News

OpenAI releases GPT-5.6 after a White House-requested delay over cyber risks

OpenAI released its GPT-5.6 models to the public on Thursday 9 July after earlier holding back the rollout over US government concerns about cybersecurity, The Guardian reports. The Trump administration had asked last month that the release be limited to a small group of government-approved users, and the wider launch came after additional testing by the government’s Center for AI Standards and Innovation, according to Axios.

The Star in Malaysia, carrying an AFP report, set out the line-up: Sol, the new flagship; Terra, a mid-range model for everyday work; and Luna, a fast, low-cost option. In late June OpenAI had shared preview access only with a limited group of US partners at Washington’s request. The Star noted that GPT-5.6 and Anthropic’s Mythos models have drawn concern over their reported ability to find software vulnerabilities that hackers could exploit.

“GPT-5.6 Sol, along with Terra and Luna, will launch publicly this Thursday (July 9).” — OpenAI, on X, via The Star

How each outlet framed it

• The Guardian — the policy angle: the staggered release follows similar US restrictions on Anthropic’s latest models, which led to a temporary export ban, and a Trump executive order last month calling for a voluntary government review of new models.
• The Star — the product: three models, a US-only preview, and a reported White House approval after tests and meetings.
• NewsBytes — the process: OpenAI sent a technical team to Washington for weeks of reviews and, citing unnamed reports, floated giving the government a small equity stake during talks on new AI rules.

What we don't know yet: what the government’s tests checked, or whether the review process, described by The Guardian as voluntary, will apply to every future model.