Anthropic's new artificial intelligence model, Claude Mythos Preview, was the center of discussion at the spring meetings of the World Bank and the International Monetary Fund in Washington.
The meeting was attended by finance ministers, heads of central banks, and other regulators, according to the Financial Times. According to the publication, senior financial officials warned that new AI models like Mythos could pose a threat to the global banking system by revealing vulnerabilities in the cyber defenses of financial institutions.
An Unexpected Agenda
According to the FT, a week ago, participants in the meetings were expecting discussions about the war in the Middle East, tensions in the private lending market, and rising government debt. However, the key topic was the ability of new AI models to destabilize the banking system.
"This is a very serious challenge for all of us," said Andrew Bailey, the Governor of the Bank of England and the President of the European Union's Financial Stability Board. "It reminds us of how quickly the world of AI is evolving." Bailey added that global regulators need to quickly assess the potential cybersecurity risks posed by the new Mythos model.
"The evolution of digital technologies poses enormous risks to cybersecurity," said IMF Deputy Managing Director Dan Kats. "This will be a key item on the international agenda in the coming months."
A New Threat
According to participants in the Washington meetings, Mythos is changing the balance of power between cybercriminals and financial institutions' security systems, as it can automatically identify and combine software vulnerabilities on a scale beyond human capabilities.
"They [global regulators] are both impressed and concerned about Mythos' ability to self-determine its approach to problem-solving, but they are also concerned about the lack of regulation for AI companies. Officials have stated that they are considering collaborating with technology companies, including Anthropic," the FT notes.
In early April, the U.S. Treasury Department held a closed meeting with the heads of major banks, including Citigroup, Morgan Stanley, Bank of America, Wells Fargo, and Goldman Sachs, as well as with the head of the Federal Reserve, Jerome Powell. The purpose of the meeting was to warn banks about the risks associated with Mythos and similar models, so that they could better protect their banking IT systems, according to Reuters.
"AI is our friend"
At the moment, Anthropic has provided access to the model to about 40 companies, including Amazon, Apple, and JPMorgan, so that they can test the system and fix the vulnerabilities that have been identified, according to the FT. JPMorgan CEO Jamie Dimon noted that the model "identifies many vulnerabilities that need to be fixed."
Some banks also emphasize the potential benefits of working with Mythos. "AI is our friend. This is a new generation of technology that will become part of the ecosystem," said Ted Peake, head of Morgan Stanley (quoted by FT). According to him, the bank is already testing the beta version of Claude Mythos and exploring where the technology can improve infrastructure efficiency.
Robin Vince, CEO of BNY, the oldest bank in the US, told the FT that the bank is also testing the Anthropic model, while Citigroup’s CFO Gonzalo Lucetti told reporters in mid-April that the bank is approaching this technological breakthrough “with the degree of urgency it requires.”
Europe is on standby
At the same time, Europe is not yet aware of the magnitude of the threat posed by the new AI models.
Senior executives at a major bank and a major financial infrastructure company in Europe told the Financial Times on the sidelines of meetings in Washington that they had not yet been given access to the beta version of Mythos and therefore had not been able to fully test it.
“If what they’re saying is true, it’s certainly serious. But it’s too early to tell. We’re in touch with Anthropic to get access to this model,” a senior bank executive told the newspaper on condition of anonymity.
François-Louis Michaud, head of the European Banking Authority, said that the assessment process is just beginning. He emphasized that the regulator is already strengthening its cybersecurity efforts and expanding its oversight of banks' IT providers. "I'm not saying that everything is perfect, but we're not starting from scratch," he said.