Deepfake era makes proving authenticity harder than spotting fakes, fraud expert warns #

A photograph of ailing US Senator Mitch McConnell became an unlikely teaching moment at the world’s largest anti-fraud gathering this month, as a digital forensics expert warned that fraud prevention is entering a phase where confirming something is real has become harder than identifying what is fake.

Dr. Hany Farid, a professor at the University of California, Berkeley and co-founder of deepfake detection firm GetReal Security, used the McConnell episode as a live case study during his keynote address at the 37th Annual ACFE Global Fraud Conference in Boston. The conference, running from July 12 to 17, 2026, draws more than 5,500 anti-fraud professionals from around the world.

The controversy arose when a photograph of Senator McConnell, released around July 12 showing the former Senate Republican leader during a hospital stay, drew waves of scepticism on social media. Users variously claimed the image was AI-generated or had been recycled from McConnell’s 2023 hospitalisation. The claims spread rapidly despite no concrete evidence supporting them.

Farid’s team at GetReal Security examined the image forensically, scrutinising faces, lighting, shading, and structural elements. They also ran the photograph through AI image-verification tools developed by OpenAI and Google. According to reporting by Poynter, the analysis found no evidence of artificial generation. Farid noted that the lighting appeared consistent with authentic photography and that neither McConnell nor his wife, Elaine Chao, who appeared alongside him, showed signs of digital manipulation.

PolitiFact, which independently fact-checked the claims, similarly concluded the photograph was genuine, finding no earlier copy of the image and no markers of AI synthesis.

For Farid, the episode illustrated a shift in the challenge facing fraud examiners. As generative AI tools grow more sophisticated, bad actors are increasingly exploiting what researchers call the “liar’s dividend”: the ability to cast doubt on real evidence simply by claiming it was artificially produced. Fraud prevention professionals can no longer assume that proving something is genuine will be straightforward, Farid argued.

The ACFE conference features more than 90 educational sessions and more than 100 speakers and has made AI and deepfake literacy a focus of its 2026 programme. Farid’s research has previously been used by governments, media organisations, and law enforcement agencies to authenticate digital media and expose manipulated content.

His keynote comes as AI-enabled fraud losses are escalating globally, with voice cloning and synthetic imagery being deployed in financial scams, identity fraud, and disinformation campaigns.

Source: Poynter