Perpetrators of AI Sexual Abuse Often View Their Actions as a Joke, New Research Shows

Why It Matters

The prevalence of AI‑driven sexual deepfakes signals a growing digital abuse risk, demanding urgent legal and platform‑level responses to protect victims and curb normalization.

Key Takeaways

• 3.2% of adults admit creating or sharing AI sexual deepfakes.
• Men, younger adults, non‑white and disabled respondents show higher perpetration rates.
• 18% deliberately view deepfakes, mainly out of curiosity.
• 26% of sharers aim to damage the target’s reputation.
• Deepfake sharing more common for men than women (56% vs 41%).

Pulse Analysis

The new study of 7,231 adults in the United States, United Kingdom and Australia reveals that AI‑generated sexual deepfakes are not a fringe phenomenon. Overall, 3.2 % of respondents reported creating, sharing, or threatening to share non‑consensual synthetic sexual images, with the highest rate in the U.K. (6.1 %). Men, younger users, non‑white participants and people with disabilities were disproportionately represented among perpetrators. The survey also found that 1.4 % engaged in non‑AI‑based manipulation, highlighting a broader culture of digital sexual abuse.

Motivations clustered around experimentation, bragging and, strikingly, reputational sabotage. Over a quarter of creators and sharers admitted they wanted to destroy the target’s reputation, while a tenth cited financial gain. Viewers, comprising 18 % of the sample, were driven primarily by curiosity, followed by sexual arousal and amusement. Gendered emotional responses were stark: men reported amusement or arousal, whereas women expressed empathy, sadness and disgust. The research underscores that the harm extends beyond the fake content, affecting victims’ mental health, privacy and professional standing.

These findings push policymakers to treat synthetic sexual media as a serious legal issue. Legislators in several jurisdictions are already drafting statutes that criminalize non‑consensual deepfake distribution, but the study argues that technical safeguards alone will not curb consumption. Sociotechnical solutions—such as platform‑level detection, user‑education campaigns, and rapid takedown mechanisms—are needed to counter the normalization of viewing and sharing these images. While self‑reporting bias may understate prevalence, the data provides a crucial baseline for future regulation and interdisciplinary research.