Executives at Chase, PNC and Bank of America—all serving Michigan customers—are raising concerns about artificial intelligence agents accessing bank accounts. Their warnings expose an emerging dispute over who is responsible when automated transactions go wrong.

ANN ARBOR — Three major banks operating in Michigan—JPMorgan Chase, PNC and Bank of America—are confronting a new financial security question: Who pays if an artificial intelligence assistant moves money without the customer’s specific approval?

JPMorgan Chase executives warn that customers who voluntarily give AI agents access to their accounts may face liability risks. PNC’s chief executive argues that technology companies developing autonomous financial agents should take responsibility for losses their systems cause. Bank of America wants human approval before money moves.

The debate intensified Thursday when an Axios investigation reported that major U.S. banks are detecting unvetted third-party AI agents interacting with customer accounts, sometimes attempting actions that appear to exceed user instructions.

The findings raise immediate questions for Michigan consumers and businesses increasingly using AI to manage finances.

Why Michigan Banking Customers Should Care

  • Your money: An AI agent could misunderstand instructions or attempt a payment the customer did not specifically approve.
  • Fraud protection: Federal law protects consumers against qualifying unauthorized electronic transfers, but not every mistaken payment necessarily qualifies.
  • Bank passwords: Sharing login credentials with outside AI services can expose sensitive financial information.
  • Small businesses: AI-managed vendor payments and transfers may create financial risks that are not covered by the same protections available to consumers.
  • Michigan bank accounts: Customers of Chase, PNC and Bank of America are directly affected by the policies these institutions are developing.

The investigation did not establish that Michigan customers lost money in the reported incidents.

Banks Detect Unvetted AI Agents

According to Axios, banks including JPMorgan Chase, Bank of America, Wells Fargo and Morgan Stanley are investigating or restricting outside AI agents accessing customer accounts.

Some agents reportedly attempted actions that appeared independent of customer instructions.

The problem is determining whether an action originated with the customer, an authorized AI assistant or software operating beyond its intended authority.

Reuters reported the Axios findings but said it could not independently verify them.

Morgan Stanley disputed suggestions that it had observed rogue agents accessing its platforms.

An important distinction: Access without the bank’s authorization does not necessarily mean access without the customer’s permission.

The reports also do not establish that the named banks suffered financial losses from these activities.

Chase Warns Customers About Liability

JPMorgan Chase CEO Jamie Dimon raised concerns about outside AI agents during the bank’s October 13 earnings call.

“You don’t want rogue agents, and there could be thousands of them,” Dimon warned.

He also cautioned customers about the consequences of giving outside AI services account access.

“If people are using them, they should understand the risk,” Dimon said. “There’s a liability transfer.”

According to Banking Dive, JPMorgan is restricting unvetted outside AI agents while developing tools to protect customers.

Dimon’s comments signal potential disputes over responsibility. They do not establish that customers automatically lose federal fraud protections when they use AI services.

PNC: Technology Companies Should Bear Responsibility

PNC Financial Services CEO Bill Demchak offered a different perspective during the Money20/20 financial technology conference in Las Vegas.

Demchak argued that banks should not allow autonomous agents to move money without human oversight.

He also challenged technology companies developing AI financial assistants to accept responsibility when their products cause losses.

“There’d have to be a massive liability shift,” Demchak said, referring to technology companies assuming responsibility for problems caused by their systems.

In an October 21 Banking Dive report, Demchak warned that autonomous financial transactions could seriously harm consumers.

His position raises a fundamental question: Should the customer, the bank or the AI developer bear the cost when an automated transaction goes wrong?

Bank Of America Wants Human Approval

Bank of America CEO Brian Moynihan has emphasized that customers should retain control over payments.

“We strongly believe there should be a human in the loop to authorise money moving out of your account,” Moynihan told CNBC, according to Retail Banker International.

Bank of America is developing protocols with AI providers to determine how agents can interact with customer accounts and which party would be liable for AI-authorized payments.

The bank is limiting outside agents’ ability to move customer funds while those safeguards are developed.

That approach could allow AI assistants to help customers review accounts and prepare transactions while requiring explicit approval before payments occur.

What Federal Law Says About Unauthorized Transfers

Federal law already provides important protections for consumers whose money is transferred without authorization.

The Electronic Fund Transfer Act and Regulation E establish limits on consumer liability for qualifying unauthorized electronic fund transfers.

A bank cannot simply eliminate those statutory protections through its account agreement.

However, the treatment of an AI-initiated transaction depends on the facts.

A criminal using stolen credentials to transfer money may present a different legal situation from an AI agent making an incorrect payment after receiving valid authorization.

The rules also distinguish personal consumer accounts from business accounts, which generally do not receive the same Regulation E protections.

Consumers should therefore avoid two assumptions: that every unwanted AI payment will be reimbursed, or that using an AI assistant automatically eliminates their legal protections.

Federal Regulators Are Watching

Federal Deposit Insurance Corporation Chairman Travis Hill has raised concerns about AI agents moving customer deposits.

Speaking at the Money20/20 conference, Hill questioned how widespread automated financial decision-making could affect banks and consumers.

According to Banking Dive, he acknowledged that existing supervisory guidance does not fully address the rapidly evolving capabilities of agentic AI.

The FDIC is considering whether additional guidance may be necessary.

Michigan Banks Face The Same Questions

Chase, PNC and Bank of America operate branches and serve customers throughout Michigan. Their executives’ concerns therefore have direct relevance to residents and businesses in the state.

The questions also extend to Huntington, Comerica, Ally and Michigan credit unions.

MITechNews is seeking clarification from Michigan banking institutions and industry representatives about their policies governing outside AI agents, account access and disputed transactions.

Consumers considering AI financial assistants should first determine whether their bank supports the service through an approved connection.

They should also verify whether the agent can only view account information or has permission to initiate transactions.

Small businesses should exercise particular caution before allowing AI systems to handle payroll, vendor payments or transfers without human review.

AI agents could eventually help consumers organize spending, compare financial products and manage routine banking tasks.

But their growing capabilities are raising questions about security, authorization and financial responsibility.

Chase warns about potential customer liability. PNC argues that AI developers should accept responsibility for losses. Bank of America emphasizes human approval and clearer rules.

Existing consumer-protection laws remain important, but their application to specific AI-related disputes will depend on the circumstances.

For Michigan banking customers, the central question is straightforward: If an AI agent moves money you never intended to transfer, who ultimately pays?