TL;DR:

• The UK government is in active talks with Anthropic over expanded Claude Mythos access for British banks and major organisations, but Anthropic has declined to commit to a timeline.
• JPMorgan chief Jamie Dimon has privately warned a UK banking executive that Mythos use should be co-ordinated with government; the Bundesbank’s Joachim Nagel has called publicly for all institutions to have access to avoid misuse.
• The UK AI Security Institute previewed Mythos before release and said the model can exploit vulnerabilities that would take human professionals “days of work”, with Technology Minister Liz Kendall and Security Minister Dan Jarvis warning cyber capabilities are accelerating faster than envisaged.

Anthropic remains in active talks with the UK government over a broader rollout of Claude Mythos to British banks and major organisations, the Financial Times reports, a week after the model first became accessible to a select group of UK financial institutions.

The state of negotiations

Officials are discussing expanded rollout to key UK organisations, people familiar with the matter told the FT. Banks and financial institutions in particular are seeking expedited access to strengthen cyber security — the use case Anthropic has designed Mythos around, given the model’s ability to identify software vulnerabilities faster than human researchers. Anthropic told the FT it would not commit to a timeline. The current US-exclusive initial tranche of 40 organisations includes Amazon, Microsoft, JPMorgan Chase and Morgan Stanley; the UK government is the only non-US state to have had a preview via the AI Security Institute.

UK financial services executives have, in the meantime, been swapping intelligence with US counterparts that already have access. One UK executive told the FT they had been discussing specific Mythos-exposed vulnerabilities with American companies, and groups were in discussions with Microsoft to source “patches” — technical fixes — so UK firms could bolster defences before formally receiving Mythos access. It is a quiet asymmetry: UK banks can learn what vulnerabilities Mythos has found without running the model themselves.

Governance concerns widen

JPMorgan chief Jamie Dimon has privately warned a UK banking executive that Mythos should be used only in co-ordination with government, per sources briefed on the conversation. Germany’s central bank chief Joachim Nagel called this week for all institutions to have access to Mythos, citing the need for a “level playing field” — and warned the model was “double-edged”, useful for defence but also for exploitation. UK Technology Minister Liz Kendall and Security Minister Dan Jarvis issued a joint public letter last week warning AI cyber capabilities are “accelerating even faster than had been previously envisaged” and urging businesses to fortify defences.

The UK AI Security Institute has publicly said Mythos can exploit vulnerabilities “that would take human professionals days of work”. Anthropic is also separately investigating reports that users gained unauthorised access to Mythos through third parties — a governance incident rather than a model-capability one, but one that hardens regulator caution over who exactly is using the tool.

Looking Forward

For UK banks, the strategic picture is now clear: Mythos access is not a switch that flips on one day but a negotiated, gradual rollout shaped by cross-market governance discussion. The Bank of England and UK Finance’s cross-market operational resilience group met this week to discuss Mythos and wider AI cyber risks, and agreed firms will use AI “to strengthen cyber defence”. The next milestone is whether Mythos access extends meaningfully beyond the initial tranche — and whether the UK government’s negotiating hand is strengthened by the AI Security Institute’s preview role.