One in four unconcerned by sexual deepfakes created without consent survey finds (www.theguardian.com)

A police-commissioned survey of 1,700 people found one in four respondents either see nothing wrong with or feel neutral about creating and sharing sexual deepfakes of non-consenting people (13% “nothing wrong,” 12% neutral). The Crest Advisory report also found 7% had been depicted in an intimate deepfake (only 51% of those reported it), one in 20 admitted making deepfakes, more than 10% said they would make one in future, and two-thirds have seen or might have seen a deepfake. Senior police and victim-support figures warned that cheaper, more accessible AI tools are normalizing sexualized deepfakes — predominantly targeting women — and that tech firms are complicit; creating non-consensual sexually explicit deepfakes is now a criminal offence under the Data Act. For the AI/ML community this underlines urgent technical and policy priorities: reliable detection and provenance tools (watermarking, robust classifiers), stronger platform moderation workflows, privacy-preserving design, and clearer auditability and reporting paths so victims aren’t silenced by shame or disbelief. The survey also flagged a demographic pattern—younger men more likely to accept or produce deepfakes—suggesting targeted education and behavioral interventions. In short, the story is a call to combine technical safeguards, legal enforcement, platform responsibility, and public education to prevent misuse as generative models become faster and cheaper.