Tavus’ Griffin video bot claim triggers scam and psychosis concerns

AI company Tavus says its new video call bot, Griffin, has passed a video Turing test, a claim that quickly drew scrutiny and safety concerns. In an October 1 post on X that drew more than 13 million views, Tavus shared footage of Griffin speaking with founder Hassaan Raza and said 48% of people who interacted with it live believed they were talking to a real human. The company also said previous systems had a pass rate below 3% and described Griffin as the first Human Interaction Model, or HIM, while claiming it ranked No. 1 on NVIDIA’s benchmark for full-duplex AI video. The claim has already been flagged on X with a community note saying the results were not independently verified and were not based on a standard protocol. According to the report, Tavus said the test involved 54 participants recruited through an independent research platform, though it did not disclose demographic details. The article says Tavus reported that among 26 people, 48% believed they were speaking to a real person, while 28 identified it as AI, with most becoming suspicious in under 20 seconds. Critics said a system like Griffin could be used to scam older people or intensify AI-related mental health risks. Protos also cited past lawsuits involving ChatGPT and Gemini, and said Tavus appears to acknowledge the safety issues by stating Griffin will need safety considerations before any public release. At the time of writing, Griffin appeared to be available only to researchers.