Bank of America has assigned 3,000 employees and 1,000 outside contractors to remediate software vulnerabilities identified with AI systems including Anthropic’s Mythos, according to comments from CEO Brian Moynihan reported by CNBC.
The work began in the second quarter. Moynihan said the bank spent a few hundred million dollars in the third quarter on the software effort and anti-money-laundering requirements together, but did not separate the amount attributable to vulnerability repair. He expects patching costs to decline over time.
Anthropic introduced Mythos to selected partners through Project Glasswing in April and said it had found thousands of serious vulnerabilities. OpenAI followed with its Daybreak cybersecurity program in May. Bank of America is among the first large US companies to disclose the workforce assigned to dealing with findings from this newer generation of security models.
The disclosure demonstrates the operational cost of turning automated discoveries into fixes, but it leaves important questions unanswered. The bank did not state how many alerts it reviewed, what share were confirmed, how severe they were or how much remediation would have occurred without AI. The 4,000-person figure therefore measures the response effort, not the precision or independent effectiveness of any one model.