Geoffrey Hinton suggested broadcasting deliberately fabricated videos, then revealing at the end that they were false and did not come from the campaign they appeared to represent. He described the tactic as a possible way to prepare audiences for video disinformation. Newtral.es published its interview report on October 21, 2026, and placed the conversation at the Starmus festival in Tenerife.
How the proposal relates to prebunking
Prebunking means alerting people to a manipulation tactic before they encounter a misleading claim. Newtral.es connected Hinton’s idea to this approach and also reported a separate field experiment: 5.4 million YouTube users saw ads about manipulation techniques, and recognition of those techniques increased by 5%.
That experiment involved ads, while Hinton suggested broadcasting fabricated videos followed by a disclosure. The reported increase was an outcome of the ad-based experiment, not of Hinton’s proposed broadcasts.
What video-authenticity checks can tell you
A visual glitch or an automated detector result alone cannot establish where a video came from or whether its claims are true. The National Institute of Standards and Technology’s November 2024 report says no single content-transparency technique provides a comprehensive solution. Provenance information about a video’s origin, labels, automated detection and human review can offer complementary signals, with usefulness depending on context and implementation.
The U.S. Government Accountability Office’s March 11, 2024 assessment describes limits to deepfake detection in real-world conditions. It also notes that identifying a fake may not stop disinformation from spreading. A visual clue or detector result is one signal, not a verdict on a video’s origin or claims.
When real footage is dismissed as fake
Newtral.es described the “liar’s dividend”: awareness of manipulated media can give people a way to cast doubt on authentic recordings by claiming they are fake.
