Until recently, the biggest concern for someone using a credit card abroad was that their bank might block the transaction as potential fraud.

The fix was to call customer service and confirm the transaction by voice. The card would be unblocked. The process relied on a simple assumption that the person on the phone was who they claimed to be.

That assumption no longer holds.

A new PLoS One study found that artificial intelligence has reached a point where cloned voices are indistinguishable from genuine ones. In the experiment, participants were asked to tell human voices from AI-generated ones across 80 samples. Cloned voices were mistaken for real in 58% of cases, while human voices were correctly identified only 62% of the time.

“AI-generated voices are all around us now,” said lead author Dr. Nadine Lavan, senior lecturer at Queen Mary University of London, per a LiveScience report. “We’ve all spoken to Alexa or Siri, or had our calls taken by automated customer service systems. Those things don’t quite sound like real human voices, but it was only a matter of time until AI technology began to produce naturalistic, human-sounding speech.”

In practical terms, AI has passed its first Turing test in sound. Alan Turing’s benchmark proposed that a machine could be considered intelligent when a human could no longer tell it apart from another human. That threshold now applies to voice. What we hear can no longer be taken as proof of who is speaking or who is authorizing a transaction.

Advertisement: Scroll to Continue

The Human Voice as a New Fraud Vector

OpenAI CEO Sam Altman warned over the summer that AI has defeated the voice ID systems many banks still rely on, calling it “a crazy thing to still be doing.” Altman told central bankers that AI can now replicate a customer’s voice with near-perfect precision and that the next phase will involve video indistinguishable from live calls.

Altman’s comments highlight a weakness in financial security. Many authentication systems are still built on the assumption that sound is a reliable indicator of identity. However, that assumption is becoming outdated.

The PYMNTS Intelligence report “The Impact of Financial Scams on Consumers’ Finances and Banking Habits” found that fraud has become targeted and adaptive.

“Scammers have adopted advanced techniques for their financial scams, drawing parallels to how businesses personalize consumer outreach,” the report said. “This shift reflects a growing sophistication in fraud strategies. Rather than one-size-fits-all targeting, savvy scammers exploit consumers’ unique circumstances.”

For banks and wealth managers, the PLoS One study exposes a growing operational and reputational risk. A voice call that sounds authentic can lead to unauthorized transfers, approval of false instructions, or exposure of sensitive information.

“As scammers continue to refine their tactics, [financial institutions] must invest in advanced analytics and behavioral monitoring to stay ahead of these tailored threats,” the PYMNTS Intelligence report said.

The Collapse of Auditory Trust

Regulators have begun to respond.

The Federal Trade Commission said in August that there has been a more than fourfold increase in impersonation scams since 2020, with losses in the hundreds of millions of dollars.

In 2024, the FTC held a Voice Cloning Challenge to address the harms of AI-enabled voice cloning technologies. The goal was to “foster breakthrough ideas on preventing, monitoring and evaluating malicious voice cloning.”

“A strong approach to AI-enabled voice cloning ensures that AI companies releasing tools that have the potential for misuse may be held liable for assisting and facilitating illegal activity in certain circumstances if they do not implement guardrails to prevent it,” the FTC said in an April 2024 press release.

The Federal Communications Commission ruled in February 2024 that AI-generated voices in robocalls violate the Telephone Consumer Protection Act.

Meanwhile, Consumer Reports Advocacy documented in August rising consumer concern, with more than 75,000 signing a petition urging stricter enforcement against voice-cloning scams.

The World Economic Forum identified in July deepfake speech and identity impersonation as emerging risks to digital financial infrastructure.

A June arXiv preprint highlighted inconsistencies in AI voice systems across languages and accents, underscoring uneven exposure for global institutions.

For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.