UK banks offered OpenAI cyber tool after Anthropic access delay
Nine banks have been offered access to GPT 5.5 Cyber, as fierce rival Anthropic has blocked previews of its tool.
By BBC News
Major UK banks have reportedly been offered access to OpenAI’s GPT-5.5 Cyber model after access to Anthropic’s rival Mythos tool stalled.
Major UK banks have reportedly been offered access to an advanced OpenAI cybersecurity tool after access to a rival Anthropic model stalled.
The BBC reported that nine banks have been offered access to GPT-5.5 Cyber, a specialist AI model designed for defensive cybersecurity work.
The offer comes after UK banks were unable to access Anthropic’s Mythos model, which has attracted attention because of its ability to identify software vulnerabilities.
Reuters previously reported that Bank of England governor Andrew Bailey said British banks still lacked access to Mythos, despite Anthropic being willing to share the model on a trial basis.
He said the delay appeared to involve political issues in the United States.
The issue matters because banks and financial firms rely on complex, older software systems that can contain hidden weaknesses.
Advanced AI cyber tools could help security teams test systems, find vulnerabilities and strengthen defences before criminals exploit them.
But the same capability also creates concern.
If powerful cyber models can find weaknesses faster than human teams, regulators want to make sure access is controlled and used only by trusted defenders.
The UK Government has already warned business leaders that Anthropic’s Mythos model showed a major increase in cyber capability when tested by the UK AI Security Institute.
The Government said the model was “substantially more capable at cyber offence” than previous systems assessed by the institute.
That explains why access to these tools is being handled carefully.
For banks, the appeal is clear. AI models trained for cybersecurity could help identify flaws in code, analyse suspicious behaviour, check patches and support teams dealing with constant cyber threats.
For regulators, the challenge is balancing defensive benefits against the risk that similar tools could be misused.
OpenAI’s GPT-5.5 Cyber has been described as a model aimed at approved cybersecurity teams and defensive workflows.
Technology reports say it is intended for work such as vulnerability detection, triage, patch validation and malware analysis, with access limited to vetted users.
For Cheshire businesses, the story may sound remote, but the practical message is important.
The same cyber risks that affect banks also affect smaller companies, schools, councils, care providers and local employers.
Many organisations rely on old systems, shared passwords, weak access controls, unpatched software or third-party suppliers.
AI will make cybersecurity faster on both sides. Defenders will be able to detect problems more quickly, but attackers may also use similar tools to test weaknesses.
That means businesses should not wait for advanced AI tools before improving basic security.
The essentials still matter: multi-factor authentication, regular software updates, strong backups, staff training, access control and a clear incident response plan.
The fight between OpenAI and Anthropic over cyber AI also shows that artificial intelligence is moving beyond chatbots and office productivity.
The next stage is likely to involve highly specialised tools for finance, healthcare, law, software development and security.
For the financial sector, access to these systems could become a competitive and regulatory issue.
For everyone else, it is another sign that cybersecurity is becoming an AI-driven race.