OpenAI Steps In to Support UK Financial Institutions
OpenAI has offered nine major UK banks access to its advanced cyber security AI model, GPT-5.5 Cyber. The offer comes as rival AI company Anthropic continues to limit access to its own cyber security tool, Claude Mythos.
Both AI systems help organizations detect hidden security flaws in software, digital infrastructure, and online platforms. Experts say these models can identify vulnerabilities faster than many traditional security methods and, in some cases, outperform human analysts.
Anthropic’s Mythos Raises Security Concerns
Claude Mythos attracted significant attention when Anthropic introduced it earlier this year. The company claimed the AI discovered a security weakness in a legacy system that had reportedly remained undetected for nearly three decades.
The announcement sparked concern among financial leaders, regulators, and banking executives. Many worry that highly capable cyber security AI tools could expose weaknesses across financial systems if they fall into the wrong hands.
As a result, access to Mythos has remained restricted. Anthropic has not made the technology widely available, particularly to major banking institutions seeking to test and strengthen their own systems.
Bank of England Calls for Greater Access
Last week, Bank of England Governor Andrew Bailey highlighted the issue during public remarks. He noted that UK banks still cannot access Claude Mythos to assess the security of their digital platforms and applications.
Financial institutions argue that access to advanced cyber security tools is essential as cyber threats become more sophisticated. Banks want to use these technologies to identify vulnerabilities before cybercriminals can exploit them.
OpenAI Presents an Alternative
George Osborne, the former UK Chancellor who now serves as a senior executive at OpenAI, responded to concerns surrounding access to advanced cyber security AI.
According to Osborne, OpenAI does not intend to keep GPT-5.5 Cyber exclusively for internal use. He emphasized that the company wants organizations to benefit from the technology and improve their cyber defenses.
However, Osborne also confirmed that GPT-5.5 Cyber will not be available to everyone. OpenAI plans to control access carefully, similar to Anthropic’s approach with Mythos.
AI Security Institute Finds Comparable Performance
The AI Security Institute recently evaluated both GPT-5.5 Cyber and Claude Mythos. Its findings showed that the two models performed at a similar level across a range of cyber security tests.
Researchers said both systems demonstrated strong capabilities in detecting vulnerabilities and analyzing complex security challenges. The results suggest that competition between OpenAI and Anthropic could drive further innovation in AI-powered cyber defense.
Growing Role of AI in Cyber Security
As cyber threats continue to evolve, advanced AI tools are becoming increasingly important for governments, banks, and businesses. While these systems offer significant benefits, experts believe careful oversight will be necessary to balance innovation with security risks.
The debate over access to powerful cyber security AI models is likely to continue as organizations seek stronger protection against emerging digital threats.
