Mythos a Serious Threat, But More Will Follow, Barclays CEO Warns
Barclays CEO C. S. Venkatakrishnan has warned that Anthropic’s advanced AI model Mythos could become a serious cybersecurity threat to the global banking system due to its powerful coding and vulnerability-detection capabilities. Speaking during IMF spring meetings, he said Mythos can identify weaknesses in legacy banking systems and potentially help create exploit paths faster than human teams, making older financial institutions especially vulnerable. He also cautioned that this is only the beginning, warning that “Mythos 2” and “Mythos 3” will likely arrive with increasing frequency, creating a long-term cyber arms race between attackers and defenders. Regulators in the U.S., U.K. and Canada are already reviewing the model, while banks are being pushed to strengthen defenses and prepare for a future where AI becomes a core cybersecurity battlefield rather than just a productivity tool.
Anthropic’s advanced AI model Mythos has emerged as a major concern for the global banking system, with Barclays CEO C. S. Venkatakrishnan warning that the cybersecurity risks posed by the model could be only the beginning of a much larger challenge for financial institutions worldwide.
Speaking on the sidelines of the IMF spring meetings at a gathering of the G30 consultancy group, Venkatakrishnan said Mythos represents a serious issue for banks because of its unusually powerful coding capabilities. According to him, the model’s ability to identify cybersecurity weaknesses and even devise ways to exploit them creates a potentially unprecedented threat to banking systems, especially those running on older legacy infrastructure.
He warned that the problem will not stop with this single model.
“But here’s the thing: there will be a Mythos 2 and a Mythos 3 and they’ll come up with probably distressing frequency,” he said, suggesting that increasingly powerful AI systems could trigger a long-term cyber arms race between attackers and defenders.
Why Banks Are Worried About Mythos
Mythos, developed by Anthropic, is considered one of the company’s most advanced frontier AI systems for coding and autonomous agentic tasks. Experts say its strength lies in its ability to rapidly scan complex digital systems, detect hidden vulnerabilities and potentially generate sophisticated exploit paths much faster than human teams.
This creates a particular challenge for banks because many large financial institutions operate a mix of modern software and decades-old legacy systems. These older systems often contain weak points that are difficult to detect and patch quickly.
Venkatakrishnan noted that larger and older institutions may face the greatest pressure because adapting legacy infrastructure is far more difficult than securing newer systems.
Regulators Are Moving Quickly
The rise of Mythos has triggered urgent discussions among regulators and financial supervisors across multiple countries. Governments in the United States, Canada and the United Kingdom have reportedly met with major banks to assess the potential risks.
Selected organizations are currently reviewing the model to understand its real-world cybersecurity implications. Regulators fear that if such AI systems are misused, they could destabilize critical financial infrastructure.
Even Andrew Bailey, Governor of the Bank of England, recently described Mythos as posing major cybersecurity risks to the banking sector and stressed the need for urgent evaluation.
A New Cybersecurity Arms Race
Barclays’ CEO believes the banking industry must now shift from simply reacting to cyber threats toward actively preparing for AI-powered attacks.
“We have to understand its capabilities and we have to understand how to safeguard against it,” he said.
This means stronger investment in defensive cybersecurity, faster patching systems and deeper collaboration between regulators, banks and AI developers.
Some institutions, including major global banks, are already participating in controlled testing programs such as Project Glasswing, where Mythos is being evaluated for defensive cybersecurity purposes rather than broad public use.
Bigger Than Just One AI Model
The larger concern is not Mythos alone, but what comes next.
As AI models become faster, smarter and more autonomous, the risk is that cybercriminals or even state-backed actors could eventually use similar tools to scale attacks across industries.
For banks, where trust and operational continuity are everything, even a small breach can create systemic consequences.
Venkatakrishnan’s warning reflects a growing reality across global finance: AI is no longer just a productivity tool it is becoming a core cybersecurity battleground.
And for the world’s financial institutions, preparing for “Mythos 2” may matter even more than defending against Mythos itself.
