OpenAI clears US security review, sets Thursday launch for GPT-5.6 model family
The release follows weeks of government testing and marks a shift in Washington’s oversight of frontier AI systems, with rival Anthropic also regaining access to its most advanced models.
OpenAI will make its GPT-5.6 model family publicly available on Thursday, the company confirmed on Tuesday, after the US Department of Commerce lifted the staged-release restrictions it had imposed in June. The decision, reported by Axios and confirmed by multiple outlets, follows technical testing by the department’s Center for AI Standards and Innovation and meetings between company engineers and officials in Washington. The three-tier release—Sol, the flagship model; Terra, a mid-range option priced at half the cost of its predecessor; and Luna, a fast, low-cost variant—had been limited to a small group of vetted US partners since late June, at the administration’s request.
The government’s intervention stemmed from concern that GPT-5.6 and similar systems, including Anthropic’s Mythos series, possess an unprecedented ability to identify software vulnerabilities that could be exploited in cyberattacks. US officials, viewing the technology through a national-security lens, sought to assess the risk that the models could be misused by military or intelligence services in China, Russia, or other countries. The review was conducted under a voluntary framework established by a June executive order, which allows developers to submit frontier models for up to 30 days of evaluation before wider release. Anthropic’s Fable 5 and Mythos 5 models were similarly restricted and then restored last week after the company implemented additional safeguards.
The launch intensifies an already heated commercial race. OpenAI’s Terra model undercuts the pricing of its own GPT-5.5, a move analysts in London and San Francisco see as a direct response to competition from Anthropic and Google, both of which are preparing their own frontier releases. Elon Musk’s SpaceXAI also announced on Wednesday that it would release its Grok 4.5 model publicly, describing it as an “Opus-class model, but faster, more token-efficient and lower cost.” Both OpenAI and Anthropic have confidentially filed for initial public offerings targeting valuations near $1 trillion, raising the financial stakes of each product cycle.
With the US government now drawing up formal criteria for which models will be subject to future security restrictions, the episode establishes a pattern of pre-release scrutiny that is likely to become routine. The immediate milestone is Thursday’s global availability of GPT-5.6, which will test whether the safeguards satisfy regulators and whether the pricing strategy can shift market share in a field where the gap between capability and cost is narrowing rapidly.
| Sub-Saharan African press | −0.40 | critical |
|---|---|---|
| Arab Gulf press | −0.30 | critical |
| Latin American press | +0.20 | neutral |
The US government's approval of GPT-5.6's public release is a dangerous gamble that prioritizes corporate interests over national security. The model's unprecedented ability to find software vulnerabilities makes it a weapon for hackers, and Washington's oversight is insufficient.
By framing the model's capability as a direct threat to national security and highlighting the potential for exploitation by hackers, the narrative creates a sense of urgency and distrust towards the US regulatory process.
The bloc omits the fact that the US government conducted additional testing and oversight before approval, as reported by other blocs. It also downplays the potential benefits of the model.
The release of GPT-5.6 is a global security concern, as its power to uncover software flaws makes it a tool for cyberattacks. Washington's approval does not mitigate the danger; rather, it exposes the world to new threats.
By emphasizing the model's unprecedented vulnerability detection and linking it to potential hacker exploitation, the narrative universalizes the risk, making it a concern for all nations, not just the US.
The bloc omits the details of the US government's additional testing and oversight framework, as well as the competitive context between OpenAI and Anthropic.
The US government has given its seal of approval to GPT-5.6 after rigorous testing, ensuring the model meets safety standards. The launch is a step forward for AI innovation, backed by proper oversight.
By focusing on the government's approval process and the new oversight framework, the narrative legitimizes the launch as safe and regulated, downplaying security concerns.
The bloc omits the initial US freeze and the national security fears raised by other blocs, as well as the concerns about vulnerability exploitation.
Broaden your view
Trump Declares Iran Ceasefire Over, Yet Agrees to Continue Talks
6 languages · 31 outlets
From Economy & MarketsSK Hynix Prices Record $26.5bn US Listing, Pre-Market Points to 21% Jump
5 languages · 13 outlets
From Science & HealthCarney’s Saudi Visit and Iran Overture Signal Canada’s Trade-First Pivot
2 languages · 5 outlets