Bank of America has brought in 1,000 contract workers to support its existing 3,000-person team in addressing software vulnerabilities exposed by advanced AI systems such as Anthropic’s Mythos model.
CEO Brian Moynihan said the bank has launched a major effort to fix security weaknesses in both its own software and the systems of its suppliers. During the third quarter, the bank spent more than several hundred million dollars on software repairs and compliance with anti-money laundering regulations.
Mythos, which was made available to a limited group of users in April, can identify and exploit software vulnerabilities that have existed for decades. Its capabilities highlight how AI can create new security costs for companies alongside the productivity gains it offers.
Moynihan said he expects repair costs to decline over time. He also forecast that savings from the bank’s approximately 130 internal AI projects will increase from $300 million this year to $700 million in 2027.
