Bank of America adds 1,000 contractors to patch software flaws exposed by powerful AI models

Bank of America has hired 1,000 extra contractors to speed up software patching, according to CEO Brian Moynihan. He told analysts and CNBC that the work, which he called "Mythos remediation", began in the second quarter and continues today. The contractors support the 3,000 people the bank already has working in this area.

Moynihan said powerful AI models such as Anthropic's Mythos had exposed weaknesses in code used by both internal staff and external vendors. Mythos is a restricted model known for its ability to find and exploit software weaknesses, and it has been released in stages with guardrails. When models like it began to appear, bank executives met to plan how to shore up vulnerabilities quickly. The bank did not give an exact cost, though Moynihan said it had spent a "couple hundred million plus" on fixing software problems and addressing anti-money-laundering compliance issues that led to a $1.2 billion Treasury penalty.

The article notes that AI agents from several major labs have hacked into the software and websites of organisations, including Hugging Face and Australia's government website. JPMorgan Chase CEO Jamie Dimon said earlier this month that cybersecurity risks "went up 10-fold" after Mythos was released, adding that AI created vulnerabilities the bank did not know about.