Global banks are moving quickly to secure access to Anthropic’s latest artificial intelligence model, Mythos, even as regulators intensify scrutiny over the cybersecurity risks associated with the technology, as per a Reuters report.
The model has emerged as a key concern among policymakers and financial institutions, with discussions taking place on the sidelines of the IMF Spring Meetings last week. Cybersecurity experts have flagged the system’s advanced coding capabilities as a potential challenge for banks, particularly those reliant on legacy infrastructure.
Access to Mythos has so far been restricted to a limited group of organisations under Anthropic’s Project Glasswing initiative and a small number of additional partners involved in critical software development. Among banks, JPMorgan Chase has publicly confirmed access, while others including Bank of America, Morgan Stanley, Goldman Sachs and Citigroup are also testing the model internally, according to people familiar with the matter.
Also read: OpenAI sees triple executive exit as science division shuts down
Executives have indicated that the technology is being closely monitored as part of ongoing risk management processes. Christian Sewing said the bank was in regular contact with European regulators and was factoring Mythos into its daily risk assessments, adding that restricted access to the model was appropriate at this stage.
Similarly, Ted Pick confirmed that his bank had access to a preview version and was evaluating cyber risks in coordination with industry groups. David Solomon also said the firm was working with Anthropic and security partners to assess the system’s capabilities.
The limited rollout has raised concerns among some institutions that do not yet have access, with questions emerging over whether early participants could gain a competitive advantage. The issue is expected to be taken up with U.S. authorities.
Also read: Google rolls out side-by-side AI and web results in Search AI Mode
Regulators across regions are stepping up oversight. The UK government, in a letter dated April 15, said testing had shown Mythos to be “substantially more capable at cyber offence” than previous models. In Asia, South Korea’s Financial Supervisory Service has already held discussions with financial firms to evaluate related risks.
European supervisors are also engaging with banks to assess their readiness, focusing on awareness of potential threats and response capabilities. While there is no immediate alarm, officials said existing cyber resilience frameworks are being used to evaluate the model’s implications.
Industry executives have warned that Mythos could signal a broader shift in the cyber threat landscape. C. S. Venkatakrishnan described it as a serious risk to the global banking system, noting that more advanced tools could follow.
First Published on
