Bank of America CEO Brian Moynihan praises AI firms’ focus on safety

Brian Moynihan, the chairman and CEO of Bank of America, has called it “encouraging” that AI companies are treating safety as a genuine priority, not just a talking point. His comments come as executives across Silicon Valley and Wall Street grapple with the uncomfortable reality that AI capabilities are advancing faster than the guardrails meant to contain them.

Leaders from Anthropic, OpenAI, and Google have been actively discussing the formation of a new industry safety body, with some even floating the idea of deliberately slowing the pace of development to let safety protocols catch up.

BofA’s AI footprint keeps expanding

Bank of America has become one of the most aggressive deployers of AI in the financial sector, with employees now generating over 400,000 AI prompts daily across the organization.

The bank has approved more than 300 AI use cases, including 114 specifically for generative AI. It operates 270 AI and machine learning models that the bank says have produced measurable gains in both productivity and security.

BofA reports a 55% reduction in fraud losses attributed to its AI systems, along with a 20% boost in developer productivity.

Then there’s Erica, the bank’s AI-powered virtual assistant. The tool interacts with 110 internal systems and can field roughly 700 types of customer questions.

The ‘human in the loop’ philosophy

Moynihan has been insistent on what BofA calls a “human in the loop” approach. The idea is simple: let AI handle the heavy computational lifting, but keep a human being involved whenever high-stakes judgment calls are required.

Moynihan raised the security dimension during a Bloomberg Television interview in July, specifically mentioning Anthropic’s Mythos as an example of the kind of rapidly evolving model that forces banks to rethink their defensive posture. His point was that financial institutions need to respond quickly to each new generation of AI capabilities, because the threat landscape shifts every time a more powerful model hits the market.

A broader industry reckoning

In September, leaders from Anthropic, OpenAI, and Google convened to discuss establishing a dedicated AI safety body, a signal that even the companies racing to build the most powerful models recognize the current regulatory framework isn’t keeping pace.

Some participants in those discussions proposed intentionally decelerating AI development until safety standards can be formalized and widely adopted.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.