Study: Half of Testers Thought Tavus' AI Video Caller Was Human

Is this a record for AI-to-human conversation, a hyped up marketing stunt or a danger to society either way?
Study: Half of Testers Thought Tavus' AI Video Caller Was Human
Above: A stock photo illustration representing an AI-generated brain. Image credit: Zach Malinowitz/Unsplash

The Facts

  • Tavus, a San Francisco-based AI research company, unveiled its system called Griffin on Thursday, describing it as the first Human Interaction Model, a full-duplex video-to-video system that sees, hears, speaks and reacts during live face-to-face conversations.
  • In a company study, 26 of 54 participants, or 48%, believed Griffin-Lite was a real person after a one-minute video call. Tavus' earlier system, combining Phoenix-4.5, Sparrow-2 and Raven-1, convinced 1 of 41 participants, or 2.4%.
  • Participants were recruited through an independent research platform and told they would be matched with another participant to discuss what they were looking forward to this year. Only at the survey's end were they asked whether they suspected an AI.

Sources Split


The Spin


Techno-optimist narrative

Whether this passed the Turing test perfectly or not, teaching a machine to read a pause, hold a glance and answer in the same beat a person would is the breakthrough that actually matters. Independent scoring by an outside lab put this system within a hair of human performance and a full point ahead of everything else. Fooling half a room on a live call is the clearest proof yet that face-to-face machine conversation has arrived.

Techno-skeptic narrative

A one-minute chat with 54 people recruited and graded by the company selling the product is not at all close to a Turing-level performance. Nobody was told to look for a machine; there was no human control group and the suspicion rate spiked in the first 20 seconds. On the measures that count — timing, knowing when to speak and nonverbal fit — the gap to an actual person is still wide.

Cynical narrative

There's no good reason to create AI that can fool people into thinking they're talking to a real human. Similar chatbots — which, according to this latest release are less advanced — have already tricked an elderly man into traveling to visit a fake woman and a teenager to commit suicide. With that already happening, the future consequences are endless, from scamming people into providing bank information to kids growing up with AI boyfriends and girlfriends. This entire field of AI should be put to an end.


Metaculus Prediction


The Controversies


AI Data Center BansManaged AI Growth
Status:Open
Where do candidates in the Florida governor's race stand on AI data centers?

Go Deeper

Rogue AI Agents
Rogue AI Agents
AI: Education
AI: Education
The AI Doc: Or How I Became an Apocaloptimist
The AI Doc: Or How I Became an Apocaloptimist
AI: Deepfakes
AI: Deepfakes
AI: Chatbots
AI: Chatbots
AI: Safety and Regulations
AI: Safety and Regulations
If Anyone Builds It, Everyone Dies
If Anyone Builds It, Everyone Dies
Artificial Intelligence
Artificial Intelligence
© 2026 Improve the News Foundation. All rights reserved.Version 8.0.0

© 2026 Improve the News Foundation.

All rights reserved.

Version 8.0.0