Bank of America CEO Brian Moynihan said the bank added 1,000 outside contractors in the second quarter to fix software vulnerabilities uncovered by advanced AI models such as Anthropic's Mythos. They work alongside about 3,000 internal staff. Software remediation together with anti-money laundering compliance work cost more than two hundred million dollars in the third quarter. The bank expects these costs to fall and projects $700 million in savings from 130 internal AI projects.

What it means: This is one of the first times a major bank has put concrete numbers on the cost of defending against vulnerabilities exposed by new AI models. The conclusion is that AI does not only bring savings, it also creates a sudden bill for anyone with old code and many software vendors. Greek banks and payment providers face a similar problem on a smaller scale: decades-old systems, third-party dependencies and limited security staff. Under DORA, responsibility for flaws in third-party software does not pass to the vendor, so contracts need fix deadlines and audit rights. For Greek software firms selling to banks, it means they will be asked to prove they have tested their code with the same tools attackers use. The right response is not panic but an application inventory, prioritised critical systems and scheduled AI-assisted reviews before rushed mass hiring becomes necessary. Security budgets for 2027 should reflect this new workload.

Read the original on CNBC

Related service: myDATA integrations

Related stories