Zelda Williams calls out a 13-second AI deepfake of her father

What circulated

The clip appeared on Elon Musk’s social network. In it, an AI-generated Robin Williams says he has worked in the industry for a long time and has “seen some things,” before claiming that the real elite — people at the very top — do not merely play games of power but worship something much darker.

Zelda Williams said the supposed private video was obviously AI-generated and that the quality was poor. One fan who believed it, she argued, seemed to be listening to her father’s films without sound: the voice was robotic and unconvincing.

A community note identified the footage as a deepfake falsely presented as a “private recording of Robin Williams from 2013.” Thousands of users still shared it. It is not possible to establish immediately whether this was the exact video Zelda was referring to, but the clip spread widely enough to prompt her response.

What Zelda Williams said

Williams addressed the people circulating such material in a separate post. Her point was that a person still sits behind the output: someone instructed a system to generate the video. She asked users not to pull her father into the “nonsense” of people detached from reality and to let him rest.

She also argued that:

Williams separately criticized the condition of Musk’s microblogging platform, saying it had become known for spreading disinformation and hate speech. She called the app a dump and wrote that it now mostly consisted of bots and people willingly falling for bots’ tricks. She did not want to return to explain herself again, she said, but her love for her father made her do it.

The detail the posts leave unresolved is provenance. The community note can label the clip, but it does not identify who generated it, why this particular conspiracy was put in Williams’s voice, or how many viewers accepted it as authentic before seeing the correction.

This is not Zelda Williams’s first objection

Williams has opposed AI imitations of her father before. In 2023, she said AI models used to imitate his voice were especially upsetting to her.

Two years later, as text-to-video generators became popular, she again asked users not to send her videos of Robin Williams. She said she did not want to watch them and would not respond sympathetically, however much senders might hope she would.

I think the significance here is less the quality of this particular fake than the permission it appears to give its audience: a dead performer can be made to endorse any belief, then left to his family to disprove. The robotic voice is a flaw in the artifact, not a safeguard against the claim.