Bank of America Hired 1,000 People to Fix What AI Found: The Hidden Cost | inicIA

Direct answer

On October 14, 2026, after reporting third-quarter results, Bank of America CEO Brian Moynihan told CNBC the bank added 1,000 external contractors in the second quarter to patch software vulnerabilities discovered by frontier AI models, including Anthropic's Mythos. The contractors are still working today, on top of the ~3,000 employees the bank already had on the task: 4,000 people fixing what AI found. Third-quarter spending on software remediation and anti-money-laundering was "a couple hundred million plus." Mythos, Anthropic's most cybersecurity-focused model, was released in April 2026 to a small set of banks, tech companies, and government agencies under controlled access; it can scan code and flag weaknesses decades old. The bank expects spending to decline once urgent patching completes. In parallel, BofA runs ~130 internal AI projects projected to deliver $700 million in savings by 2027 (versus $300M this year) — and cut ~3,400 jobs, though without directly linking it to AI.

The two faces of the finding

What your company must budget (the part nobody quotes you)

  1. Remediation, not just detection: when buying AI tools for security, code, or auditing, ask who fixes what they find and what it costs. A finding without remediation capacity is just a worry list.
  2. Technical debt in the light: AI exposes years of shortcuts in your code and processes. Do the inventory before a model (or an attacker with a model) does it for you.
  3. The complete ROI: BofA projects $700M in savings from its 130 AI projects by 2027 — but this year it pays hundreds of millions in remediation. AI ROI is calculated net of its security costs, not just on savings.
  4. Patch speed as a KPI: if your patch cycle is measured in weeks, you are already behind attackers using AI. Measure and shorten that cycle.

Related reading

Frequently asked questions

What is Anthropic's Mythos?

It is Anthropic's most cybersecurity-focused model: it can analyze code and find vulnerabilities. Since April 2026 it has been available to a small, vetted set of organizations (banks, tech companies, agencies), under a tiered cyber-verification program.

Why did Bank of America need 1,000 contractors?

Because Mythos and other frontier models found vulnerabilities in its code — some decades old — faster than its internal team could patch them. The contractors were added in Q2 2026 and remain active.

Does this mean AI destroys more jobs than it creates?

The case shows both at once: BofA cut ~3,400 positions (without directly attributing it to AI), hired 1,000 temporary specialists, and projects $700M in savings from 130 AI projects by 2027. AI reallocates work more than it simply eliminates it.

What is the lesson for a mid-size company?

That AI's cost does not end with the license: it includes fixing what AI discovers in your systems. Budget remediation from day one and measure your patch speed.

Take the next step

AI will show you everything that is broken in your systems. Better you discover it — with a plan to fix it — before an attacker does. inicIA helps you deploy AI with the real costs on the table. WhatsApp: https://wa.me/17866120930

Chat on WhatsApp

By inicIA