Banks' answer to the AI alarm: "We just don't do that"

Banks' answer to the AI alarm: "We just don't do that"

The statements on Monday, September 14, 2026, came in the wake of a week in which Anthropic's chief executive asked the AI industry to slow down — and the two banks, which themselves use AI heavily, chose to position themselves within the safety debate, not against it.

Two of America's largest banks stepped into the debate over AI risk on Monday, September 14, 2026: Bank of America's Brian Moynihan assured audiences that the bank never lets AI agents operate without supervision, while PNC's president Mark Wiedman made agent oversight a top priority.

The week that preceded it

The backdrop is a week marked by unease in the AI industry. Anthropic researcher Jacob Coxon resigned the preceding Tuesday with warnings about a rush toward self-improving superintelligence, and Anthropic's chief executive Dario Amodei published an essay arguing that AI companies "must slow the pace at which we improve the capabilities of AI models." According to American Banker, three other AI producers — Sam Altman at OpenAI, Demis Hassabis at Google DeepMind, and Elon Musk — expressed agreement the same day. It is against this backdrop that Moynihan and Wiedman chose to speak (American Banker, source 6d7adc39).

Moynihan: No autonomous agents

At the Barclays Global Financial Services Conference, Moynihan acknowledged the week's warnings directly. "This weekend was a lot about the risk in AI, and that largely is around agents just left to operate," he said, according to American Banker, adding: "We just don't do that" (source 6d7adc39).

Moynihan also explained the mechanism behind the bank's approach: BofA, according to him, decided long ago that employees should always check everything AI produces for customers — that is, human review as standing practice, not a measure introduced in the wake of the week's debate (source 6d7adc39).

The same day, Bloomberg TV covered an interview in which Moynihan called it "encouraging" that AI companies are taking responsibility for AI safety seriously, even as leaders sound the alarm about the pace of innovation and its possible consequences. The remark was given to Bloomberg's Dani Burger (source 144d1004). The supplied video material says nothing about exactly where or in what context the statement was made beyond the interview itself, so the two statements cannot be definitively assigned to different venues.

Wiedman: Agent oversight as a top priority

PNC president Mark Wiedman, who leads the bank with $616 billion in assets, struck the same note. Referring to the public debate of recent days, he said, according to American Banker, that ensuring agents do what they are supposed to do is "a top priority for us" (source 6d7adc39).

However, the American trade publication reproduces Wiedman's statement broken off mid-sentence, so the full scope of PNC's position is not available in the material at hand. What is verifiable reporting is that he explicitly ties agent oversight to the public debate of the past week.

How big are the banks' AI operations?

There are figures suggesting why the banks are engaged. Crypto Briefing writes, citing Bank of America itself, that the bank has approved more than 300 AI use cases, of which 114 are for generative AI, and operates 270 AI and machine learning models (source 7e3fec14). These figures are not corroborated by other sources in this material and are based on the company's own statements — they should be read as the bank's self-description, not as independently verified facts.

There are also no primary sources for the debate itself: no transcripts of Moynihan's conference remarks, the Bloomberg interview, or Wiedman's comments have been provided. All quotes rest on secondary reporting from American Banker, Bloomberg, and Crypto Briefing.

What remains?

What is clear after September 14: Two banks that use AI heavily chose to respond to the industry's safety alarm by positioning themselves within its framework — not against it. Moynihan praises the AI companies for their focus on safety, but stresses that BofA has never unleashed agents without supervision; Wiedman makes agent oversight a top priority for PNC.

What remains unresolved: Whether these practices are as the executives describe them cannot be verified by readers from the available sources — the claims rest on the executives' own statements, as reported in secondary sources. And the question Amodei raised about how fast model development itself should proceed touches the banks only indirectly: They are AI users, not model producers, and their answer does not settle the debate over the pace of capabilities — it only shows how some of the largest customers in the AI economy now expect safety from the supplier side.