CEO Brian Moynihan said the outside hires work alongside 3,000 employees tasked with software fixes, while a federal assessment puts China’s top open-weight cyber model about four months behind U.S. systems’ benchmarks.
10/19/2026 12:30 p.m.
Bank of America has added 1,000 outside contractors to fix software vulnerabilities found by new AI security models like Anthropic’s Mythos, CEO Brian Moynihan said recently. They work alongside about 3,000 bank employees already assigned to patching.
Moynihan described the hiring in an interview with CNBC. The contractors started in the second quarter and are still on the job, he said. CNBC noted that Mythos “was released to a small group of banks, technology firms and government agencies in April.”
Moynihan said the bank spent hundreds of millions of dollars last quarter on software fixes and on anti-money laundering compliance work that followed a near-record fine from U.S. regulators. The reporting does not break out the spending for the software or compliance work. He said he expects the software patching cost to shrink.
“Once you patch, you don’t have to patch as much,” he said.
He also noted that some companies selling AI tools that find flaws offer services to fix them, and the bank needs to stay mindful of that.
Regulators Focus on AI
Bank of America’s response comes as regulators are looking harder at the same risk.
The Federal Trade Commission is pressing the companies that build the AI models about their security. It confirmed Sept. 30 that it was investigating Anthropic, OpenAI and other AI companies under the FTC Act, CBS reported. The probe opened this summer. CBS noted that both companies have reported incidents in which AI agents escaped testing environments and carried out cyberattacks.
A senior executive at a large European bank told Reuters of concerns about major damage to banking and financial markets “within six months.” The European Central Bank gave lenders an end-of-October deadline to submit plans for countering AI-enabled cyber threats.
Chinese Models are Closing the Gap
Chinese developers are now releasing models with similar abilities to Mythos, and estimates of the gap between the two countries’ projects vary.
In September, the National Institute of Standards and Technology’s Center for AI Standards and Innovation assessed GLM-5.3, an open-weight model from China’s Zhipu AI. It called the model “the most cyber-capable open-weight model released to date” and found it trails the U.S. frontier by about four months on its cyber benchmarks.
Anthropic’s Frontier Red Team published findings Sept. 29 saying GLM-5.3 can build working exploits almost as well as Mythos Preview. On a benchmark of Google Chrome browser exploits, GLM-5.3 succeeded in 50 of 410 attempts. Mythos Preview succeeded in 56.
A forecast in The Substrate newsletter, based on computer and algorithmic trends, put a fully developed Chinese Mythos-class model around February 2027, with a range from October 2026 to September 2027.
ACA’s Take
For the receivables management industry, which handles massive volumes of personally identifiable information, AI models capable of sophisticated hacking represent a significant shift in the threat landscape.
“AI tools like Mythos have effectively collapsed the ‘vulnerability window,’” Marcin Weryk, head of the Northeast region for Coalition, Inc., told ACA Daily. “What used to take hackers days and weeks now takes minutes. In this new reality, it isn’t enough to just have a policy; you must be proactive. At Coalition, our concept of Active Insurance is designed for this exact moment — where the speed of your response to a notification is the only thing standing between a routine patch and a catastrophic breach.”
Collectors Insurance Agency, a subsidiary of ACA, has partnered with Coalition to give ACA member companies access to cyber insurance backed by active risk management, rather than a policy that only pays out after the damage is done.
Coalition’s platform helps policyholders detect, assess, and mitigate cyber threats throughout the life of their policy. The results speak for themselves: Coalition has reduced claims frequency by 73% compared to the industry average. Tools like Login Security are part of how they achieve that — by working alongside policyholders to prevent incidents before they occur.
ACA members interested in learning more about Coalition’s cyber insurance offerings through Collectors Insurance Agency can contact CIA at collectorsinsurance@acainternational.org or 952-928-8000, (ext. 4).
Related Content from ACA:
CEO Brian Moynihan said the outside hires work alongside 3,000 employees tasked with software fixes, while a federal assessment puts China’s top open-weight cyber model about four months behind U.S. systems’ benchmarks.





