Tavus Defends AI in Courtroom After Avatar Jim Appears on Behalf of Dewald

On March 26, a New York appellate courtroom became the unlikely stage for a scene that could have been lifted from a legal tech thriller. At the heart of the drama stood 74-year-old Jerome Dewald, representing himself in an employment dispute against MassMutual Metro New York. Dewald, a throat cancer survivor, had received approval to present his oral argument via video. But no one, not even the presiding judges expected what came next.

The screen flickered on, revealing not Dewald, but a slick, articulate man in a V-neck sweater. “Now may it please the court, I come here today a humble pro se before a panel of five distinguished justices…” the figure intoned. It wasn’t Dewald. It wasn’t even a real person. It was “Jim,” a synthetic avatar built using AI video tech from a startup called Tavus.

“Hold on,” interjected Associate Justice Sallie Manzanet-Daniels, eyebrows raised. “Is that counsel for the case?”

“That?” Dewald replied, sitting calmly in court. “I generated that. That is not a real person.”

The judge was stunned. “You did not tell me that, sir. You are not going to use this courtroom as a launch for your business.”

_{Source: The Register}

That business is Pro Se Pro, an AI-driven legal aid startup Dewald created to help unrepresented litigants navigate the court system. He said the avatar wasn’t a marketing stunt, it was a workaround for his medical condition. “Extended speaking is difficult for me,” Dewald explained later. “I did get permission in advance. They were just unprepared to see an artificially generated image.”

But the issue wasn’t just the tech. It was the delivery. Dewald originally intended to use an avatar that looked like himself, but his budget plan with Tavus allowed for only three renders a month. The first two failed to process. “So I just used one of their stock replicas —that big, beautiful hunk of a guy they call Jim,” he said with a shrug.

The judges didn’t see it that way. The video was halted just seconds in. Dewald was forced to deliver his statement live—the very thing he had hoped to avoid.

“AI Is Here. The Question Is How.”

Tavus has now come in the limelight and AIM Research spoke to Quinn Favret, Co-founder and COO of Tavus, the reaction though understandable missed the broader picture.

“We believe the judge’s reaction came from a place of surprise as she was caught off guard,” he said in an interview following the hearing. “When new technology enters high-stakes environments like the courtroom, surprise and skepticism are natural, although we always hope it receives benefit of the doubt.”

Disclosure, he stressed, is key to building trust. And in this case, it wasn’t made explicitly clear to the court that the video featured an AI-generated avatar. “The technology made this possible,” said Favret. “Jerome used an AI-generated video because he has speech challenges and wanted to present a professional and audible statement in court. If the judge had known, she may have been more receptive.”

_{Source: The Register}

Favret emphasized that the situation wasn’t about deception, but about access, something Tavus is deeply focused on. “Ultimately, we see AI avatars as an extension of and for the empowerment of humans—not as a replacement for,” he said. “We see a world where AI makes legal systems more accessible—especially for pro se litigants who might not otherwise get a voice.”

Ethics and Deepfakes

Still, the reaction in the courtroom raised real questions. If a judge can’t tell that an avatar is synthetic, what’s to stop the rise of deepfakes in legal proceedings?

Favret doesn’t shy away from the concern. “We approach responsibility on three levels: technology, culture, and regulation,” he explained. “On the tech side, we’ve built guardrails to prevent impersonation, such as multi-step verification to ensure people can’t clone others without consent.”

He added: “We actively advocate for disclosure mechanisms to support clearer boundaries in sensitive environments like the legal system.”

Tavus avatars are built with transparency features baked in. “For example, when you ask one of our real-time video agents who they are, they always disclose that they are AI,” Favret noted. “In our experience, users appreciate that transparency. Once they know, they welcome interacting with AI.”

The courtroom may have made headlines, but Tavus sees broader potential. “Our platform has already helped people with speech impairments, social anxiety, or mobility challenges participate in conversations they otherwise couldn’t,” said Favret. “We have customers building companions for the elderly to combat loneliness. Others are building healthcare intake agents, to bring back a human touch to care. Some are building tutors that deliver personalized learning.”

Favret is clear about the company’s stance: “We stand firmly on the side of assistive technology.”

And while he doesn’t see Tavus avatars arguing entire legal cases anytime soon, he believes the right guardrails can unlock powerful new use cases. “We believe courts should establish clear protocols: when is AI representation acceptable, what disclosures are required, and what rights are preserved or potentially jeopardized,” he said. “AI is here to stay. The question is how can we shape its use, and what role does regulation play in embracing AI and shaping this use.”

As for Jim? He’s unlikely to appear in court again any time soon. But in his brief moment on the courtroom screen, he sparked a debate that’s far from over.

“You can chat to Charlie on our website.” Favret added. “You’ll notice that Charlie can see, listen, speak, and even looks like a real human!”