In the last year, in the European Union, phishing success rate rose from 12% to 54% when comparing traditional attacks to AI-assisted attacks.
Internal AffairsEuropean Union
- Error detected
- Geographic scope error: the claim states the data applies to 'the European Union', but all identifiable sources trace these figures to US-based academic experiments, not to EU-specific cybercrime statistics.
Contextual error: the figures describe experimental click-through rates, not real-world cybercrime success rates across an entire political/economic region.
- Omissions
- The MEP attributes the figures to the European Union, but the underlying research is from US-based academic institutions (Harvard, MIT) and is not EU-specific. No EU agency (ENISA, Europol, Eurostat) has published these exact figures for EU cybercrime.
- The data refers to click-through rates in controlled academic experiments, not to real-world 'success of cybercrime by phishing' across the EU.
- The claim states 'in the last year' (implying 2025–2026), but the key studies date from 2024, and the figures represent a cross-sectional experimental comparison rather than a year-over-year trend observed in the EU.
- The MEP did not cite any source for these figures. The most frequently cited original study appears to be a 2024 Harvard Kennedy School Executive Education study or the December 2024 arXiv paper 'Evaluating Large Language Models' Capability to Launch Fully Automated Phishing Attacks' (arXiv:2412.00586).
- Sources
- AcademicarXiv (Cornell University)The control group emails received a click-through rate of 12%, the emails generated by human experts achieved 54%, the fully AI-automated emails... This paper evaluates LLMs' capability to launch fully automated phishing attacks. The study was not limited to or conducted within the European Union.
- SecondaryVectra AIAI-generated phishing achieves 54% click-through rates compared to 12% for traditional campaigns, according to 2025 research. No mention of EU-specific scope.
- SecondaryHunto AIIn a 2024 study at Harvard, LLM-generated phishing emails yielded a 54% click-through rate vs. 12% for human-written phishing emails. The study was conducted at Harvard (US), not in the EU.
- SecondaryDashlane BlogAI-powered phishing campaigns achieve a 54% success rate, compared to just 12% for traditional attacks. Does not attribute these figures to the European Union.