Bank of America has engaged 4,000 contractors to address cybersecurity vulnerabilities identified by frontier AI models. The bank began hiring these workers in the second quarter and continues to employ them to manage the remediation process.
The initiative follows findings from the Anthropic Mythos Preview model, which uncovered over 10,000 cybersecurity flaws across 50 partner organizations. Large banks with access to the model have shared findings with smaller institutions to accelerate security updates.
According to reports, the use of AI has reduced the time required to patch vulnerabilities from weeks to days. Bank of America expects its elevated spending on this effort to decrease as the identified software flaws are resolved.
Industry leaders acknowledge the dual-use nature of frontier AI, which can assist in both patching and exploiting long-standing software vulnerabilities. JPMorgan CEO Jamie Dimon noted that cybersecurity risks increased tenfold following the release of the Mythos model.
In response to rapid AI developments, the U.S. Treasury Department has initiated a series of meetings with regulators and financial institutions to coordinate security strategies. These sessions are intended to continue on a regular basis.
