LONDON — Financial fraud units across Europe and North America are escalating their response to a new generation of scams that deploy artificial intelligence voice cloning, deepfake video and hyper-personalised phishing messages to bilk consumers and businesses out of billions of dollars annually, according to a joint report released Friday by the International Consumer Protection Enforcement Network.
The report, which surveyed 34 member jurisdictions covering more than 900 million people, found that AI-assisted fraud incidents rose 61 percent in 2025 compared with the prior year, with total losses to individuals and enterprises estimated at $142 billion globally. Regulators described the figures as a conservative floor because many victims — particularly businesses — decline to disclose incidents publicly.
The mechanics of modern scams have shifted dramatically in the past two years. Fraudsters now routinely harvest publicly available voice samples from social media to generate convincing audio impersonations of family members, executives and government officials. In one documented case cited by the report, a finance director at a mid-size logistics firm in Germany wired 2.3 million euros to a fraudulent account after receiving a video call that appeared to show her chief executive authorising the transfer. Investigators later confirmed every visual element of the call had been synthesised in real time.
“The cognitive load on ordinary people is simply unsustainable,” said Dr Miriam Ochsner, a behavioural economist at the University of Basel who contributed research to the report. “These attacks are no longer mass-blast operations. They are individually targeted, emotionally calibrated and timed for moments of reduced vigilance — late evenings, travel disruption, medical stress. The human brain was not built to defend against this.”
Counterfraud coalitions are responding on several fronts. Telecommunications regulators in the United Kingdom, Canada and Australia have mandated new call-authentication protocols requiring carriers to verify the numeric origin of calls before connecting them, a framework known as STIR/SHAKEN. Early data from a 14-month pilot covering four major carriers in the United Kingdom suggest the programme has reduced successful impersonation calls by approximately 38 percent within its scope. Banks in the same pilot reported a corresponding 22 percent drop in authorised-push-payment fraud tied to phone-initiated contact.
Technology firms are also racing to develop real-time deepfake detection. Two startups — Veridian Labs and Praxis Signal — demonstrated tools at a regulators’ conference in Vienna last month that claimed detection accuracy of 94 percent for AI-generated video at standard streaming resolutions. Critics, however, note that detection efficacy degrades sharply at lower bit rates common on mobile networks and older hardware, leaving many vulnerable users unprotected.
Law enforcement agencies have intensified cross-border operations against fraud syndicates. A coordinated action in March involving police from eleven countries dismantled a network operating out of three jurisdictions in Southeast Asia that was estimated to have defrauded more than 40,000 victims across Europe and the Gulf states over eighteen months. Forty-seven individuals were arrested; asset seizures totalled approximately $380 million.
Consumer advocates argue that enforcement, while necessary, remains reactive. They are pushing for mandatory reimbursement schemes that place the primary burden of loss on financial institutions rather than victims. The European Commission is expected to publish a draft directive on the subject later this year; the United Kingdom extended its voluntary reimbursement framework to a statutory footing in January, requiring banks to compensate victims of authorised fraud up to a ceiling of 85,000 pounds per claim.
“Reimbursement is important but it does not reduce harm,” said Teodora Varga, director of digital consumer rights at the Brussels-based advocacy group SafeNet Europe. “The goal must be prevention. Every institution that touches money or personal data needs to be accountable for the security architecture it deploys.”
Analysts expect the fraud arms race to intensify through 2026 as generative models become cheaper and easier to access. The ICPEN report recommends a coordinated global standard for digital identity verification — anchored to cryptographic credentials rather than knowledge-based authentication — as the most durable long-term defence. Whether regulators in major economies can align on such a framework quickly enough to matter remains the central open question.