Bank of America has hired 1,000 contractors to patch software vulnerabilities exposed by advanced AI models, including Anthropic’s Mythos.
Bank of America has brought in around 1,000 external contractors to help patch software vulnerabilities exposed by advanced AI models, including Anthropic’s Mythos, CEO Brian Moynihan said in a television interview. According to Moynihan, the bank moved earlier in 2026 to strengthen a wide range of its internal and vendor code using thousands of employees and outside personnel. The contractors, added in the second quarter of 2026, remain at the bank alongside the 3,000 staff already working in that area. Mythos was released in April 2026 to a small group of banks, technology firms, and government agencies.
The cost of remediation
Moynihan told analysts that the bank spent more than a couple of hundred million dollars in the third quarter on software remediation and on regulatory issues related to anti-money laundering compliance. The latter follows the US Treasury’s recent penalties against the bank over anti-money laundering failures. His comments are among the first from a major US company to specify the staffing and spending involved in protecting software systems against such AI models. Moynihan described the risk as real, adding that when frontier models emerged, the bank and its peers had to work out how to address the vulnerabilities quickly.
Moynihan stated that the bank must stay ahead of these risks, while remaining aware that the companies developing the software are also the ones selling services to the bank. He also sought to reassure investors that the increase in spending is temporary, noting that once systems are patched, less patching is required, and that the bank expects remediation costs to decline over time.
AI as both a cost and a saving
AI is also reducing costs elsewhere at the bank. Bank of America runs about 130 internal AI projects, and Moynihan told analysts that savings from these efforts are expected to rise from USD 300 million in 2026 to USD 700 million in 2027. The bank’s headcount fell by about 3,400 employees over the past year, including a reduction of roughly 5% of managers, although it is unclear how much of this is attributable to AI.
The case illustrates the two sides of AI adoption in banking. The same class of models that can raise productivity can also uncover and exploit decades-old software flaws, pushing institutions with large legacy code bases and extensive vendor dependencies into significant defensive spending. For Bank of America, the cost of that remediation is currently running alongside the savings it expects AI to deliver.
