Home / Technology / AI Cyber Defense: UK Banks Get OpenAI's GPT-5.5
AI Cyber Defense: UK Banks Get OpenAI's GPT-5.5
2 Jun
Summary
- OpenAI's GPT-5.5 Cyber is now accessible to nine major UK banks.
- Anthropic's Claude Mythos, though advanced, is not yet available to them.
- Both AI systems demonstrate high performance in detecting cyber vulnerabilities.
OpenAI has granted access to its GPT-5.5 Cyber system to nine prominent UK banks. This advanced AI is designed to identify hidden weaknesses within digital infrastructure, potentially surpassing human capabilities in certain cybersecurity tasks. In parallel, competitor Anthropic's Claude Mythos model, which recently gained attention for identifying a long-standing system flaw, has not been extended this early access in the UK.
Both AI systems underwent examination by the AI Security Institute, which reported comparable performance levels in their tests. Despite this, UK lenders are currently unable to use Claude Mythos for their own system security assessments, as stated by Bank of England Governor Andrew Bailey. He has requested a demonstration of the model's capabilities to the Financial Stability Board, which he chairs.
George Osborne, a senior figure at OpenAI, emphasized the company's intention to share these powerful tools responsibly. He stated that access is intentionally limited, ensuring that AI capabilities are in the hands of those maintaining order rather than those seeking to disrupt or commit crime. While Anthropic is working to expand access to Mythos, they maintain it operates at a higher capability level requiring a more cautious rollout than GPT-5.5 Cyber.
