Bank of America brought in 1,000 external contractors to patch software vulnerabilities exposed by advanced artificial intelligence models, including Anthropic's Mythos, chief executive Brian Moynihan told CNBC on Wednesday.
Speaking after the bank released its third-quarter earnings, Moynihan told the network the company had rushed earlier this year to shore up a broad range of its internal and vendor code, drawing on thousands of its own staff alongside the outside hires. He described the risk as real, and said that once the frontier models arrived many at the bank sat down to work out how to get through the problem quickly. The contractors were added in the second quarter and are still working on the task, on top of the 3,000 people already in that area.
The patching effort, combined with regulatory work on anti-money laundering compliance, cost the bank a "couple hundred million plus" in the third quarter, Moynihan told analysts earlier in the day. His remarks are among the first from a large US company to put a figure on the staff and money spent defending software systems against Mythos, which Anthropic released to a small group of banks, technology firms and government agencies in April.
Moynihan noted that the bank had to stay ahead of the models, and that the firms building the software were often the same ones trying to sell services to the bank. With the shares falling as much as 8% on Wednesday, he sought to reassure investors that the higher spending would not last. He explained that once a system is patched, it does not need patching as much, and that the pipeline moves through, adding that remediation costs should fall over time.
Elsewhere, artificial intelligence is cutting costs. Bank of America runs about 130 internal AI projects, with savings expected to climb from $300 million this year to $700 million in 2027, Moynihan told analysts. The bank's head count dropped by roughly 3,400 over the past year as it trimmed about 5% of managers, though it is unclear how much of that reduction stems from AI.
