How deepfakes undermine video-evidence credibility
A research project is to examine whether deepfakes 鈥 AI-manipulated images, videos or audio 鈥 make courts less likely to trust evidence of human-rights violations gathered on mobile phones.
A Swansea legal expert has been awarded 鈧1.5 million to examine how public perceptions of deepfakes affect trust in user-generated evidence.
Yvonne McDermott Rees (Professor of Law at the Hillary Rodham Clinton School of Law at the University of Swansea) has been awarded European Research Council (ERC) funds as part of the EU鈥檚 research and innovation programme.
User-generated evidence, such as videos recorded by witnesses on their mobile phones, now plays an important role in legal trials worldwide, adding to knowledge of human-rights violations, and often holds perpetrators to account.
Realistic images
However, the public is also increasingly confronted with examples of 鈥榙eepfakes鈥 鈥 extremely realistic images, videos or audio recordings created using machine-learning technology 鈥 which are only likely to become more advanced and difficult to detect as the technology progresses.
Through an innovative methodology combining legal analysis of trials with mass online experiments and mock jury trials, Prof McDermott Rees's project will develop the first systematic account of 鈥榯rust in user-generated evidence鈥 (TRUE), in the specific context of its use in human-rights accountability processes.
TRUE will run from 2022-27, enabling it to track the impact of advances in technology over time.
The professor said: "Scholarship to date has expressed a concern that the rise in deepfakes will lead to mass mistrust in user-generated evidence, and that this in turn will decrease its usefulness in legal proceedings. This may well be the case, but no study has yet tested that assumption.鈥
She welcomed the ERC鈥檚 generous support in tackling 鈥渁 major evidence gap that urgently needs to be addressed鈥.