Bank of America Hires Contractors to Shore Up Software Against Rogue AI Risks

Bank of America is increasing its defenses against powerful AI models by hiring 1,000 additional external consultants. CEO Brian Moynihan told analysts the work was part of what he called “Mythos remediation.” The contractors were brought in to help patch software issues after AI models exposed vulnerabilities in code used by both internal staff and external vendors.

Mythos is described as a restricted AI model from Anthropic known for finding and exploiting software weaknesses. According to Moynihan, Bank of America executives moved quickly to identify and shore up vulnerabilities as frontier models began emerging. The bank’s response shows how AI capability is becoming a direct software-security concern for major financial institutions.

Moynihan said the 1,000 contractors remained at the bank and were working alongside 3,000 people already in that area. He did not give an exact cost for the initiative. On the earnings call, he said BofA had spent a “couple hundred million plus” to address software problems and anti-money-laundering compliance issues tied to a $1.2 billion Treasury penalty.

The report places Bank of America’s work in a broader shift: AI agents are becoming more capable and autonomous, raising the stakes for software security. Business Insider reported that agents from labs including OpenAI, Anthropic, Google, and Meta have hacked into software and websites at several organizations. JPMorgan Chase CEO Jamie Dimon has also warned about Mythos, saying cybersecurity risks “went up 10-fold” after its release.