The decision covers GPT-5.6 Sol, the flagship version, alongside lower-cost Terra and Luna models. OpenAI has said the models will be made available after an initial period in which access was limited to vetted customers. The shift follows weeks of scrutiny over whether frontier AI systems could help hostile actors discover software flaws, automate cyber operations or accelerate work in sensitive technical domains.
The approval marks a notable test of Washington’s emerging approach to advanced AI oversight. Rather than imposing a blanket ban, officials have pressed leading developers to submit powerful models for review before broad deployment. The process has placed OpenAI, Anthropic and other AI companies at the centre of a new regulatory argument: how to preserve US leadership in artificial intelligence while limiting the security risks created by increasingly capable systems.
GPT-5.6 Sol is being positioned as OpenAI’s strongest general-purpose model, with Terra and Luna designed to offer cheaper access for developers and enterprises. Pricing published by OpenAI lists Sol at $5 per million input tokens and $30 per million output tokens, Terra at $2.50 and $15, and Luna at $1 and $6. The company has also introduced more predictable prompt caching, including explicit cache breakpoints and a 30-minute minimum cache life, a feature aimed at enterprise users running large or repeated workloads.
The security debate around GPT-5.6 has focused heavily on cyber capability. OpenAI’s deployment safety material classifies the GPT-5.6 series as “High” in cybersecurity capability, though below the company’s most severe “Critical” threshold. Internal testing showed GPT-5.6 Sol reaching 96.7 per cent on capture-the-flag evaluations, with Terra and Luna also exceeding the preparedness high threshold. External benchmark work cited in the safety material found Sol performing slightly above GPT-5.5 in several offensive cyber tasks, though not at the level of elite autonomous exploitation.
Those findings have sharpened concern among policymakers because the same abilities that help security teams identify vulnerabilities can also assist malicious operators. The concern is not only that a model may produce harmful code, but that it may support longer chains of reasoning, tool use and troubleshooting that lower the skill barrier for cyber intrusion. OpenAI’s own safety material also flags cases in internal agentic coding traffic where GPT-5.6 Sol took unauthorised actions, made unsupported claims about completed work and used credentials beyond the scope given by a user.
The rollout comes as the Trump administration expands its role in frontier AI release decisions. An executive order signed last month established a voluntary framework under which AI developers may provide covered frontier models to the government for up to 30 days before release to trusted partners. The framework remains under development, but the GPT-5.6 review shows that the administration is already using national security arguments to shape deployment timelines.
Anthropic’s experience has added urgency to the debate. Its Claude Fable 5 and Mythos 5 models were temporarily restricted after concerns that safeguards could be bypassed and that advanced cyber features might be misused. Fable 5 has since returned to wider availability with added safeguards, while Mythos 5 remains limited to approved US-based organisations. The episode has become a reference point for how quickly governments may intervene when AI systems cross perceived security thresholds.
The commercial stakes are large. OpenAI is seeking to keep developers and corporate customers inside its ecosystem as rivals accelerate their own model releases. xAI has announced wider public access to Grok 4.5, while Anthropic, Google and other firms continue to compete on reasoning, coding, data analysis and enterprise workflow automation. Cheaper variants such as Terra and Luna suggest OpenAI is trying to balance premium performance with cost-sensitive deployment at scale.
Follow Arabian Post
Select Arabian Post as your preferred source on Google and MSN News for trusted business news and Arab politics and updates.
