Tavus Restricts GriffinLite After Video Call Study

Tavus says 26 of 54 people who spoke with its GriffinLite artificial intelligence system in one-minute video calls believed they were talking to a real person, a result the company cites as evidence of the model’s ability to sustain lifelike conversations. The company is limiting access to selected trusted testers while it develops disclosure features and safety measures before any wider release. The participant results come from Tavus’s own study and have not been independently verified, according to the supplied reporting. Participants were told they would meet another person for a brief conversation about what they were looking forward to that year. They were asked only at the end whether they had suspected their partner might not be real. Tavus says those who became suspicious generally did so within about 20 seconds. The study does not establish that GriffinLite passed a standard Turing test. Participants were not told that an AI might be on the call, and the test did not ask them to distinguish a hidden machine from a human under a defined protocol.