SEON, the AI command centre for fraud prevention and AML compliance, has published research showing that while American consumers are rapidly adopting AI agents, 65% believe the developer of the agent should carry the blame if it completes a fraudulent transaction.
The study polled 2,026 US adults on their habits, trust levels and concerns around AI agents. It found that 68% now rely on an agent for a personal task at least weekly, yet half would feel uneasy letting one use their credit card details to finish a purchase they had asked for.
Boomers were the most hesitant at 65%, followed by Gen X at 52%, Gen Z at 47% and Millennials at 34%. Women expressed more unease than men, at 56% against 43%.
Appetite for sharing data shifts sharply with its sensitivity. Respondents were most open to handing over location (38%), purchase history (34%) and calendar access (34%), but just 15% would share financial details. Younger groups proved more willing, though still in the minority, with 23% of Millennials and 22% of Gen Z prepared to do so compared with only 4% of Boomers. Almost a third (32%) would give an agent no information at all.
On accountability, consumers placed responsibility for agent-driven fraud firmly on AI companies, with far smaller shares pointing to themselves (14%), the merchant (11%) or their bank or card issuer (10%). The leading worries, each cited by 20%, were a criminal taking control of the agent or linked accounts, and the agent disclosing personal or financial data without consent.
Confidence in spotting deception is also thin. Some 43% doubt they could tell whether an agent’s request is genuine, and 39% do not trust agents to flag risky situations for them. Among Boomers, 61% feel little or no confidence in judging such requests, against 28% of Millennials.
Access to human help when problems arise was the strongest trust driver at 50%, ahead of the ability to undo errors (42%) and robust identity checks (39%), although 23% said nothing would raise their trust.
The firm argues that spotting agent involvement is merely the first step in judging risk, since an agent may be following a customer’s wishes or acting for a fraudster. It says firms should weigh agent activity against identity, device and behavioural data to establish who is in control, whether the action fits the customer’s permissions and whether the wider pattern signals fraud.
With products such as Meta’s Muse and OpenAI’s Dots signalling growing consumer comfort with delegated actions, SEON warns that banks, FinTechs and payment providers must prepare to distinguish legitimate agents from malicious ones.
SEON CEO and co-founder Tamas Kadar said, “Consumers are comfortable letting AI agents help with everyday tasks, but trust drops quickly when money and identity are involved. Half say human support would increase their trust, and they’ve already decided who is accountable when agent-driven transactions go wrong.”
He added, “Businesses need to tell a helpful agent from a harmful one before money moves. SEON’s real-time risk signals show who is behind agent activity, so companies can adopt agents with confidence and keep customers protected.”
Copyright © 2026 RegTech Analyst
