OpenAI offers UK banks GPT-5.5 Cyber after Anthropic blocks them from Claude Mythos

The OpenAI–Anthropic rivalry escalated into a direct contest over enterprise cybersecurity. Per the BBC, OpenAI offered nine major UK banks access to GPT-5.5 Cyber after Anthropic declined to give them previews of Claude Mythos. Both models are designed to find hidden security weaknesses in digital systems; Mythos drew attention in April when Anthropic claimed it uncovered a flaw in a legacy system that had stayed hidden for nearly 30 years.

Mythos's capabilities alarmed regulators. Several finance ministers, central bankers, and financiers expressed serious concerns that such a model could undermine the security of all financial systems, and Bank of England governor Andrew Bailey said last week that UK banks still could not access Mythos. Into that vacuum, OpenAI stepped with a competing offer — turning a regulatory bottleneck into a sales opening.

The UK's AI Security Institute, which tested both, reported they reached "a similar level of performance" on its task set, suggesting the differentiator is access and trust rather than raw capability. Anthropic's gatekeeping reflects its safety-forward positioning, but it also handed OpenAI a competitive wedge with one of the most lucrative, security-sensitive customer segments.

The episode fueled community debate over vendor gatekeeping in security-critical AI: should the most capable offensive-security models be widely available, or tightly controlled? It also separately surfaced scrutiny of ChatGPT's scale — analyses estimate ~900 million weekly active users and ~1.17 trillion prompts annually, raising fresh questions about energy consumption. Watch whether other central banks follow the Bank of England's caution, and whether Anthropic loosens Mythos access in response to the competitive pressure.