Bank of America has added 1,000 contractors to help about 3,000 in-house staff repair software weaknesses identified by advanced AI models, including Anthropic’s Mythos, CEO Brian Moynihan said Oct. 14.
The contractors joined in the second quarter and were still working with the bank when Moynihan disclosed the effort. Their work covers weaknesses in both the bank’s own code and software supplied by vendors, according to CNBC’s report. Moynihan said the bank moved quickly to address the risks as frontier AI models made it possible to uncover software flaws.
Mythos, released in April to a limited group of organizations, was among the models used; Moynihan did not specify how many flaws were found or how severe they were. Moynihan said third-quarter spending on software remediation and anti-money-laundering regulatory matters combined amounted to “a couple hundred million plus.” The bank did not break out how much of that total went specifically to patching AI-identified flaws.
He expects remediation costs to decline as the work progresses. The effort illustrates how AI can speed up the discovery of vulnerabilities while creating a substantial job—and expense—for people who must fix them.