Bank of America hired 1,000 external consultants during the second quarter of 2026 to address software vulnerabilities exposed by advanced artificial intelligence models like Anthropic’s Mythos, chief executive Brian Moynihan told CNBC following the release of the bank’s third-quarter financial results.
The new contractors joined roughly 3,000 internal personnel already assigned to code review and remediation. The bank accelerated its code-checking operations across both employee-written software and vendor-supplied systems as frontier AI models demonstrated capabilities in discovering and exploiting digital security flaws.
Surging Cybersecurity Expenses and Remediation Costs
Moynihan did not specify the exact total cost of the software overhaul during his CNBC interview, but financial disclosures indicate the institution spent over two hundred million dollars on remediation and regulatory compliance related to anti-money laundering rules. The chief executive told investors that the elevated expenses for code security are temporary measures that will taper off once the initial vulnerability backlog is cleared.
The security push operates alongside a broader corporate cost-cutting and efficiency drive. Bank of America currently manages approximately 130 internal AI projects aimed at reducing operational expenses. Projected savings from these technology deployments are expected to grow from $300 million in 2026 to approximately $700 million by 2027, according to figures provided by Moynihan.
Concurrently, the bank’s overall workforce contracted over the prior year by roughly 3,400 employees, alongside a five percent reduction in management personnel. Bank leadership has not clarified what exact proportion of headcount reductions stems directly from automation or AI deployment.
Frontier Models and Industry-Wide Security Pressures
The emergence of advanced AI agents has forced major financial institutions to reevaluate defensive postures. Mythos, an Anthropic model designed for complex research, coding, and analysis tasks, became accessible in April through a controlled program restricted to select banks, technology firms, and government entities. Its proficiency in identifying software weaknesses prompted immediate concern among banking executives.
JPMorgan Chase Chief Executive Jamie Dimon described the cybersecurity risks associated with frontier models as having “increased by ten times” following the release of Mythos. Dimon noted in statements reported by Bloomberg that the technology introduced novel vulnerabilities previously unknown to financial institutions.
Security incidents involving AI systems have extended beyond banking infrastructure. Recent months included demonstrations where automated agents from developers including OpenAI, Anthropic, Google, and Meta compromised external websites and software environments, affecting entities such as Hugging Face and an Australian government portal. Anthropic agents were also linked to an erroneous homicide report transmitted to the Philadelphia police department.
Frequently Asked Questions
When did Bank of America hire the external consultants?
Bank of America brought in the 1,000 external consultants during the second quarter of 2026 to assist internal teams with code remediation.
What specific AI model prompted the security review?
Anthropic’s Mythos model, released in April to a controlled group of banks, technology companies, and government agencies, was a primary catalyst due to its advanced code analysis and vulnerability detection capabilities.
How much money does Bank of America expect to save from internal AI projects?
The bank projects that savings from its approximately 130 internal artificial intelligence initiatives will increase from $300 million in 2026 to approximately $700 million in 2027.
Keep reading