Facephi: nearly 75% of organisations plan to adopt agentic AI, but only 21% have a mature governance framework
The Facephi observatory published "The New Fraud Frontier (Foresight 2027)" on 19 November 2026, a study that places the governance of agentic artificial intelligence at the centre of risk management.
According to the report, nearly 75% of organisations plan to use this technology within the next two years, but only 21% have a mature agent governance model.
Less than 1% of fraud losses are recovered, according to the study
The document draws on official sources, sector analyses and the Spanish publisher's own expertise. It covers six regions (North America, Latin America, Europe, the Middle East and North Africa, Asia-Pacific, Africa) and includes a five-level maturity model, ten questions for the risk committee and seven monitoring indicators.
According to the study, less than 1% of the money lost worldwide to fraud is recovered. For Facephi, the blind spot lies in the method: many organisations fail either to detect their own attacks or to trace the origin of their losses.
The report cites INTERPOL, according to which AI-based fraud is estimated to be 4.5 times more profitable than traditional fraud. It notes that in 2026, governments began treating the most advanced AI models as strategic assets, subject to export controls and "sovereign AI" strategies.
Finally, the study describes the "dormant identity": an account opened with flawless verification, sometimes based on a synthetic identity, that remains inactive for months. It is then reactivated to receive a fraudulent payment, serve as a mule account or launder funds, a type of fraud rarely recorded as such, according to the report.
For 2027, Facephi anticipates fraud extending beyond banking
For 2027, Facephi anticipates two developments. According to the study, fraud will spread beyond regulated sectors to travel, gaming, online commerce and the public sector, and a growing share of participants in digital channels will not be people but AI agents.
"Just as we now talk about account takeover, we will tomorrow talk about agent takeover," warns the report. It notes that, across the six regions studied, regulation is shifting from one-off identity checks towards continuous, evidence-based monitoring: by 2027, organisations will have to demonstrate that they prevent fraud.
The response proposed by the publisher, which specialises in facial biometrics and digital identity, consists of treating identity as a continuum: verification, session, account and transactions correlated within a single file.
"We must stop asking customers whether they are who they claim to be, and start asking whether they are still who they said they were," summarises Javier Barrachina, R&D Director at Facephi.