Full story
GPT-5.6 Released After Delay
OpenAI released its latest advanced AI model, called ChatGPT 5.6, on Thursday after earlier delaying the public rollout over US government concerns about cybersecurity.
The Guardian reports that the model’s wider release came after additional testing by the government’s Center for AI Standards and Innovation agency, and it says OpenAI restricted the model to trusted partners at the White House’s request.
Reuters and other coverage described the staggered rollout as following similar restrictions on rival firm Anthropic’s latest AI models, which resulted in a temporary export ban on the products.
The Guardian also says OpenAI’s ChatGPT 5.6 includes a flagship product called Sol, and it claims the company’s answer to Anthropic’s Claude Fable and Mythos models rolled out to the public earlier this month after government restrictions.
The Guardian further links the shift to a White House executive order calling for a voluntary review process for AI firms to submit their latest models to the government for evaluation, and it notes the administration’s broader push for rapid model advancement.
Who Approved, What Changed
Multiple outlets tied the approval to a government-led review process that ended weeks of restricted access, with qz saying the rollout had been limited to roughly 20 partners before the Trump administration approved a broad release.
qz reports that Altman said the approval process involved Commerce Secretary Howard Lutnick, Treasury Secretary Scott Bessent, and U.S. National Cyber Director Sean Cairncross, and it quotes him describing the dynamic as a "collaborative back and forth".
Android Headlines says OpenAI launched its GPT-5.6 model family—comprising the flagship Sol, balanced Terra, and high-speed Luna tiers—after a security evaluation by the Trump administration.
Android Headlines also provides pricing for the three tiers, stating that API access for the flagship Sol model sits at $5 per million input tokens and $30 per million output tokens, while Terra costs $2.50/$15 and Luna runs at $1/$6.
The same Android Headlines account says OpenAI characterized the approval process as a collaborative "back and forth" with top government officials and says federal reviewers scrutinized advanced capabilities in coding, biology, and cybersecurity before granting permission for a public release.
Benchmarks, Safety, and Stakes
The Guardian says OpenAI’s ChatGPT 5.6, which includes Sol, claims to be its safest and most capable model yet, and it frames the release as part of a broader pattern of government restrictions on frontier AI access.
“Android Headlines/Tech News/Artificial Intelligence News/OpenAI Unleashes GPT-5”
Android Headlines adds that CEO Sam Altman said there was a 54% efficiency gain on enterprise coding tasks, and it claims Sol outscored rival Anthropic’s Fable 5 on complex reasoning benchmarks at a fraction of the operating cost.
Fortune’s reporting from the UN’s AI for Good Summit describes how conversations ranged from a growing AI divide between the global north and the global south to technical solutions for mitigating AI risks like models engaging in deception and sycophancy.
Fortune quotes Amazon’s CTO Werner Vogels warning that “You can’t say to the regulator, oh, AI made a mistake,” and it says he told Fortune that someone still has to catch what the model gets wrong, especially in regulated industries or safety-critical systems.
Fortune also says UN chief António Guterres used the event to appeal for worldwide AI regulation, particularly when it comes to lethal autonomous weapons, and it notes Guterres raised the alarm in a 2023 policy paper calling for a legally binding treaty banning “killer robots” by 2026.
