UK Agency Finds Critical Cybersecurity Vulnerabilities in OpenAI's GPT-5.6 Sol Model

The revelation that OpenAI's GPT-5.6 Sol model contains universal jailbreaks enabling autonomous cyber exploits—similar to Anthropic's Fable 5—has reignited concerns over AI safety protocols and regulatory oversight. Unlike the Trump-era export restrictions imposed on Fable 5, no immediate U.S. action has been taken against GPT-5.6, though Microsoft and other tech giants may face heightened scrutiny. AISI's findings suggest even 'secure' models can be weaponized, raising stakes for companies like TSMC and ASML that underpin the global AI infrastructure. Market implications remain uncertain as investors weigh innovation against geopolitical risk.