OpenAI to Launch GPT-5.6 Amid National Security Concerns

OpenAI is preparing to publicly launch its most capable AI model, GPT-5.6, after a delay last month prompted by the US government's national security concerns. These concerns stem from the potential misuse of powerful AI systems, particularly in the context of escalating competition between the US and China to develop advanced AI technologies. Experts warn that such technologies could significantly enhance the sophistication of cyberattacks, especially in sectors reliant on outdated technology systems.

In response to these risks, Washington has intensified its scrutiny of advanced AI model releases, aiming to identify possible threats from nations like China and Russia. Concurrently, Chinese authorities have engaged with leading tech firms to discuss potential restrictions on overseas access to their most advanced AI models, which have yet to be released.

The situation has led to significant actions from OpenAI's competitors. Anthropic, for instance, temporarily disabled its advanced AI models, Mythos 5 and Fable 5, following a US export control order issued on June 12. These restrictions were lifted last week after Anthropic implemented specific safeguards.

Reports indicate that the Trump administration approved the launch of GPT-5.6 after further testing and discussions between OpenAI and government officials. Initially, access to GPT-5.6 was limited to a select group of vetted partners, whose identities were shared with authorities. In addition to GPT-5.6, OpenAI plans to introduce lower-cost models named Terra and Luna.

Elon Musk, who leads SpaceXAI, a competitor to both Anthropic and OpenAI, announced that his company would also make its leading model, Grok 4.5, available to the public. In a related move, US President Donald Trump signed an executive order that establishes a voluntary framework for AI developers to present their advanced models to the US government for review prior to public release.

While export controls for Anthropic's Fable model have been lifted, access to Mythos, which is tailored for cybersecurity professionals, remains restricted to select trusted US organizations. In China, there are concerns that the Mythos model could be used to exploit software vulnerabilities, raising fears that it could be deployed against Chinese interests. Anthropic has cautioned that achieving complete robustness against jailbreaks in AI models may be nearly impossible, highlighting the ongoing challenges in ensuring the security of advanced AI technologies.