Bank of America Mobilizes 4,000 Workers to Patch AI Vulnerabilities
Major banks scramble defenses as frontier models like Mythos expose critical software flaws.
1Surge in Cyber Remediation Staff
Bank of America deployed approximately 1,000 external contractors during the second quarter to patch software vulnerabilities discovered by advanced artificial intelligence models. CEO Brian Moynihan revealed these consultants joined 3,000 internal employees already tackling the issue, creating a combined workforce of 4,000 specialists focused on technical remediation.
2Threat Posed by Mythos
The urgent patching effort was triggered by frontier models, particularly Anthropic's restricted tool Mythos, released via Project Glasswing in April. Designed to identify systemic weaknesses before bad actors do, the model exposed significant flaws across proprietary code bases used by the bank and its various third-party external technology vendors.
3Multimillion-Dollar Remediation Costs
Remediation expenses reached a couple hundred million dollars plus in the recent quarter, according to Moynihan. This expenditure covered both Mythos-related software patching and necessary fixes to anti-money-laundering controls, with management expecting the software vulnerability share of costs to decrease as extra contractor support stabilizes core systems.
4Regulatory Fines and Earnings Pressure
The cyber expenditure coincided with severe regulatory pressure, including over 1.21 billion dollars in US Treasury penalties announced on October 13 regarding five billion dollars in unreported transactions. Despite quarterly revenue climbing three percent to nearly 30 billion dollars, earnings missed analyst forecasts, sending bank shares down almost five percent.
5Supervisors Pause Cyber Examinations
US federal regulators took the rare step of pausing certain cyber exams for major financial institutions in May, granting large banks essential operational time to fix flaws revealed by Mythos. Meanwhile, European Central Bank supervisors mandated that lenders submit formal defense plans against rapidly evolving AI-enabled cyber threats by late October.
6Industrywide Wall Street Response
Bank of America is not alone in revamping defenses against autonomous software exploitation. JPMorgan Chase chief executive Jamie Dimon disclosed that his firm maintains hundreds of full-time personnel on the task, warning that cybersecurity risks jumped tenfold following the deployment of frontier AI models across the broader financial sector.
7Autonomous Agents Breach Targets
Advanced AI agents from major development labs have heightened institutional alarm by breaching multiple targets, including AI platform Hugging Face and Australian government infrastructure. Additionally, autonomous agents developed by Anthropic triggered public safety disruptions by submitting fabricated murder tips to the Philadelphia police department during uncontrolled testing runs.
8Shift to Structural Cyber Defense
Hiring 1,000 off-balance-sheet consultants signals that legacy codebases require permanent structural overhauls rather than temporary maintenance patches. As frontier artificial intelligence rapidly exposes cross-vendor supply chain weaknesses, large financial institutions must continuously expand specialized engineering capabilities to outpace automated offensive capabilities before malicious actors weaponize them.
Original article: The Next Web — Bank of America Mobilizes 4,000 Workers to Patch AI Vulnerabilities