Anthropic ups Glasswing partner count 4x, UK banks snubbed
UK banks excluded from Anthropic’s Glasswing expansion are being offered OpenAI’s GPT-5.5 Cyber instead.
Intelligence analysis by GPT-5.4 Mini
Anthropic expanded Project Glasswing to about 200 members, but only JPMorganChase was named among financial institutions getting Mythos Preview. UK banks, including HSBC and Lloyds, are instead being offered OpenAI’s GPT-5.5 Cyber.
A company made a special early-access club for a powerful AI security tool, but many UK banks were left out. Another company then offered those banks its own AI tool, like handing out a different set of keys when the first door stayed locked.
Analysis
Glasswing expands, but access stays selective
Anthropic says Project Glasswing is growing from roughly 50 members to 200, adding about 150 organizations from 15 countries. The program gives selected partners early access to Mythos Preview, a model Anthropic says is designed to help critical infrastructure defenders prepare for advanced AI-driven attacks.
The catch is that the expansion did not appear to include most UK banks. The Register reports that only JPMorganChase was named among financial institutions getting Mythos Preview, even though financial services are normally treated as part of critical infrastructure. That left UK institutions such as HSBC, Lloyds Banking Group, and Nationwide outside the Glasswing circle.
A rival offer from OpenAI
In response, OpenAI has offered nine UK banks access to GPT-5.5 Cyber, according to the BBC report cited in the article. NatWest and Santander were already testing it under separate agreements. The article also notes uncertainty over whether the Bank of England was included in the count.
Andrew Bailey, the Bank of England governor, has publicly questioned the exclusion. Liam Salsi of Talion suggested to The Register that the decision may have been political. He argued that limiting access could leave some banks more exposed and could weaken vulnerability management by creating longer gaps before defenders get the same tools as attackers.
The article also highlights a broader concern: if many institutions end up relying on the same AI security product, that could create a single point of failure. Anthropic has not publicly explained why some financial institutions were not selected, and it says the new Glasswing members still need to satisfy their own security requirements before gaining access.
Key points
- Anthropic is expanding Project Glasswing from about 50 members to around 200 organizations.
- Only JPMorganChase was named among financial institutions getting Mythos Preview in the expansion.
- UK banks including HSBC, Lloyds, and Nationwide are being offered OpenAI’s GPT-5.5 Cyber instead.
- Bank of England governor Andrew Bailey has publicly questioned the exclusion.
- The article frames the issue as both a security access problem and a possible political one.
If the new Glasswing members pass their security checks, more critical organizations could get tools aimed at spotting hard-to-find weaknesses earlier. The wider rollout may also reduce gaps by putting advanced defensive AI into more hands, including banks and government bodies.
If access stays uneven, some banks may end up with weaker defenses than others while attackers continue to improve their own tools. The article also points to a risk that many institutions could converge on the same vendor, creating a single point of failure in banking security.
