Bank outsources the workload surfaced by AI scanning
Bank of America worked with roughly 1,000 outside contractors in the second quarter to fix software weaknesses flagged by AI models. CEO Brian Moynihan shared the figure in a CNBC interview following the bank's third-quarter results.
Moynihan said one of the models surfacing those weaknesses was Anthropic's Mythos. In this case, AI is not the attacker but a scanning layer that finds security gaps in the existing codebase.
Why it matters
A major bank bringing in hundreds of outside people to handle the remediation work generated by AI-assisted scanning points to a new bottleneck in enterprise software security: models find flaws quickly, but fixing them still takes human labor.
It also shows that AI adoption in finance brings not just efficiency but extra cost and staffing needs. Regulators asking banks about AI agents and cyber risks in the same period suggests the issue is being watched closely on the oversight side too.
