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

OpenAI moved to capture nine major UK banks by offering access to GPT-5.5 Cyber, a security tool designed to surface hidden vulnerabilities in digital systems, after Anthropic blocked those same banks from previews of its Claude Mythos security model. The episode escalates the OpenAI–Anthropic rivalry into a direct contest over security-critical enterprise AI, where vendor gatekeeping decisions now have real commercial consequences.

The BBC reporting frames a clean competitive swap: Anthropic's restriction created an opening, and OpenAI filled it. For the banks, AI tooling that can probe their own systems for weaknesses is high-value but high-sensitivity — exactly the kind of workload where access policies, liability and trust matter as much as raw capability.

Separately, analysis cited alongside the story estimates ChatGPT now reaches roughly 900 million weekly active users and processes around 1.17 trillion prompts annually, figures that have intensified scrutiny of the product's energy consumption and the 'global AI boom comes with a power bill' narrative.

The community read this as a referendum on gatekeeping in security AI. Some applauded Anthropic's caution about who gets early access to offensive-capable security models; others saw OpenAI's aggressive enterprise grab as the more pragmatic play for banks that need tooling now. The broader theme — vendors using access policy as competitive weaponry — is likely to recur as security-focused frontier models proliferate. Watch whether regulators or the banks themselves weigh in on the appropriateness of AI tools that hunt for system weaknesses, and whether Anthropic loosens or tightens Mythos access in response.