The world’s leading financial administrators have expressed concerns after a newly released Mythos artificial intelligence model from American AI company Anthropic found vulnerabilities in many major operating systems.

Mythos’ technical capabilities were uncovered in April when developers tested running the models, and their performance to undergo unethical operations was reported as “strikingly capable at computer security tasks”.

Experts warned that finance ministers, banks and people around the world should be concerned about the model’s potential to be a landmark in identifying and exploiting cybersecurity weaknesses, prompting crisis meetings.

During the test, the developers also discovered that Mythos can uncover old software bugs or create new avenues to exploit systems, which finance ministers, bank executives, and financiers said could be dangerous for financial systems.

Mythos is one of Anthropic’s latest models, developed as part of its broader AI system, Claude, a rival to OpenAI’s ChatGPT and Google’s Gemini. The company has not released the model for public use due to safety concerns.

However, Mythos has been made available to giants of the technology industry such as Amazon Web Services, CrowdStrike, Microsoft, and Nvidia as part of an “effort to secure the world’s most critical software.”

Canada’s finance minister, François-Philippe Champagne, is among those who have voiced concerns about the model, stating that the issue was discussed at the International Monetary Fund meeting this week.

“Certainly it is serious enough to warrant the attention of all the finance ministers,” Mr Champagne said. “The difference is that the Strait of Hormuz—we know where it is and we know how large it is… the issue that we’re facing with Anthropic is that it’s the unknown, unknown.”

He added, “This is requiring a lot of attention so that we have safeguards, and we have a process in place to make sure that we ensure the resiliency of our financial systems.”

A report by AI Security Institute, a UK-based research organisation that was given access to test the Mythos, also reached the same conclusion that the model is capable of exploiting undefended systems.

“Our testing shows that Mythos Preview can exploit systems with weak security posture, and it is likely that more models with these capabilities will be developed,” the report stated.

Meanwhile, Anthropic released a new version of its existing model, Claude Opus, on Thursday, saying it will allow Mythos’ cyber capabilities to be tested on less powerful systems.