The U.S. government recently imposed restrictions on Anthropic's Fable and Mythos models, raising concerns across the industry. The decision reflects heightened scrutiny of advanced AI systems that could pose national security risks.
OpenAI has begun rolling out its latest GPT‑5.6 series, which includes the flagship Sol, the mid‑tier Terra, and the efficient Luna models. These offerings map to the capabilities of Anthropic's Claude lineup and are designed to compete with the restricted models.
Sol is positioned as a direct competitor to the banned Fable and Mythos, delivering advanced coding, biology, and cybersecurity functions. The model is marketed as the company's strongest offering to date.
Unlike Anthropic's earlier general release of Fable before the ban, OpenAI has opted for a limited preview. Access is currently restricted to a select group of trusted partners and organizations.
OpenAI informed the U.S. government of its plans and the models' capabilities. The rollout is being constrained at the government's request to ensure compliance with regulatory requirements.
The company states that the limited release will facilitate a smooth transition to broader availability. It also notes that the approval process may delay access for developers, enterprises, and other stakeholders.
To address concerns about security breaches, OpenAI has incorporated layered safeguards into GPT‑5.6. These measures target jailbreaking attempts and real‑world adversarial pressure.
OpenAI anticipates that Sol, Terra, and Luna will soon become available to the general public. However, final clearance for widespread deployment remains pending.
