Fraud Losses Rarely Recovered as Agentic AI Scales

Facephi's The New Fraud Frontier report warns that AI-driven fraud is growing more profitable and harder to trace across six regions.

Updated on Nov. 19, 2026 in Artificial Intelligence

Fraud Losses Rarely Recovered as Agentic AI Scales

Live Poll

Do you trust your bank to protect your money from online fraud?

Facephi released The New Fraud Frontier (Foresight 2027) report, finding that less than 1% of global fraud losses are ever recovered and that many organizations cannot identify their own attacks or the source of their losses. The report says agentic AI can plan and execute fraud campaigns end to end.

Why it matters

INTERPOL estimates AI-powered fraud is 4.5 times more profitable than traditional fraud, and the report notes that only 21% of organizations have a mature agent governance model. The gap between adoption and oversight leaves institutions exposed as fraud expands beyond regulated industries.

Agentic AI systems can plan and execute fraud campaigns end to end without continuous human direction, according to the report. The report covers six regions, including North America, Latin America, Europe, the Middle East and North Africa, Asia-Pacific and Africa, but the extraction does not specify how the adoption and governance percentages were sampled.

The players

Facephi

A company headquartered in Alicante, Spain, that released the report and whose clients span more than 30 countries.

INTERPOL

The International Criminal Police Organization, which estimated that AI-powered fraud is 4.5 times more profitable than traditional fraud.

The details

The report describes two emerging methods: dormant identities, where accounts are opened with flawless verification, sometimes using a synthetic identity, and then stay inactive for months before use; and authorized push payment fraud, where the customer is real and approves the transaction on their own device after being persuaded by an impersonator. The authors project that fraud will spread beyond regulated industries into travel, gaming, e-commerce and the public sector.

Timeline

  1. 2026: Governments began managing frontier AI models as strategic assets subject to export controls and sovereign AI strategies.

  2. Within two years: Nearly 75% of organizations plan to use agentic AI.

  3. 2027: Organizations are expected to have to prove fraud prevention as regulation moves toward continuous, evidence-based oversight.

The Tech Race

INTERPOL's estimate that AI-powered fraud is 4.5 times more profitable than traditional fraud frames the report's warning that criminal use of agentic AI is accelerating. The report's projection that fraud will shift from account takeover to agent takeover positions defenders on a race against attackers who are adopting the same tools.

Customers of banks, travel, gaming and e-commerce services may encounter more authorized push payment scams, where they approve a payment themselves after being persuaded by an impersonator. Organizations that rely on verification at account opening should expect dormant accounts to be a growing risk.

The takeaway

Fraud checks that stop at account opening will miss attacks that wait months before acting. Reviewing activity on accounts long after they are verified is a practical step for any organization exposed to these threats.

Further reading

For broader context on how these tools are reshaping criminal activity, see Artificial Intelligence coverage.

Source note: This article includes information reported by News Makers.

Live Poll

Do you trust your bank to protect your money from online fraud?