Generative AI fraud forces shift toward platform accountability and EU regulation

The professionalization of generative AI fraud marks a transition where technical safeguards and user education are no longer sufficient to prevent financial loss. As synthetic media becomes indistinguishable from reality, the streaming and digital media ecosystem must pivot from reactive moderation to proactive verification of advertising and recommendation systems. This shift places a heavy compliance burden on large online platforms to detect anomalies before they reach the end user. The industry should monitor how strictly EU regulators enforce Article 50 transparency requirements on legitimate AI providers versus the rise of unlabelled deepfake content.

The EU AI Act's transparency obligations are becoming the primary regulatory lever against synthetic media fraud. In August 2025, the European Commission published its first draft code of practice for general-purpose AI models, which includes specific provisions requiring providers to label AI-generated content and disclose when systems produce synthetic media. Article 50 of the AI Act mandates that deployers of deepfake systems must clearly indicate that content has been artificially generated, with enforcement beginning in August 2026. This timeline directly shapes how platforms operating in the EU must prepare detection and labeling infrastructure before the compliance deadline arrives.

Platform accountability under the Digital Services Act is already generating enforcement actions that intersect with generative AI fraud. In February 2025, the European Commission opened formal proceedings against Meta under the DSA over deceptive advertising practices on Facebook and Instagram, citing systemic failures to protect users from scam advertisements, many of which increasingly use AI-generated imagery and cloned voices. The DSA's Article 26 requires very large online platforms to assess and mitigate systemic risks including manipulation and fraud, creating a legal basis for regulators to demand proactive detection of synthetic content in ad systems. Montenegro, while not yet an EU member, is aligning its digital governance framework with these standards as part of its accession negotiations, which gives the University of Montenegro's research direct policy relevance.

Technical detection of AI-generated fraud content remains an active research area with measurable progress. In March 2025, researchers at the University of Oxford published a benchmark study showing that multimodal detection models achieved 94% accuracy on deepfake video identification, though performance dropped significantly when adversarial perturbations were applied to the synthetic media. The study highlighted that audio-only deepfakes, commonly used in vishing and social engineering attacks, remain harder to detect than video forgeries, with accuracy falling below 80% for compressed audio clips. These findings underscore why the EU AI Act transparency obligations complement rather than replace technical detection, since adversarial manipulation can evade even state-of-the-art classifiers deployed at platform scale. As EU AI Act enforcement enters operational phase, platforms must also navigate new synthetic performer disclosures that are emerging in global markets, while Frankfurt court finds Meta liable for fraudulent investment ads highlights the growing legal risks for platforms.