Google rolled out a major update to Vids, its AI-powered video editor: Personal Avatars. With a selfie and a 30-second voice clip, Vids generates a digital version of you that can speak any script you type. Previously, Vids only offered 12 pre-made avatars — now it's custom to your face, voice, and mannerisms. The update also brings Gemini Omni natural language editing — describe a change and it happens. The feature went live around July 16, 2026.
Critics have already dubbed it "Google makes it easy to deepfake yourself" — and the timing is notable. OpenAI's Sora launched a "cameo" feature allowing users to appear in AI-generated social videos. Zoom CEO Eric Yuan continues pushing AI clones for meetings. The industry is normalising AI avatars faster than regulation can keep up.
The Verdict — Every Easy Button Creates an Easier Weapon
Three things can be true at once. Google Vids Personal Avatars is a genuinely useful enterprise productivity tool — it saves time, eliminates retakes, and makes async video communication frictionless. It is also, by its nature, a technology that can be repurposed for deception. And there is no evidence that Google, or any of its competitors racing to ship AI avatars, has adequately grappled with the second truth.
The concern cascade is real. If you can deepfake yourself with a selfie and voice clip, you can deepfake anyone whose face and voice are available online — executives, politicians, colleagues. A compromised Workspace account could generate a fake CEO video authorising a transfer, extending the deepfake CFO scams that cost companies millions in 2024-25. And in an enterprise setting, the authentication burden shifts: how does a team verify that a video is really their manager versus a Vids-generated avatar with stolen credentials?
Google has implemented safeguards — watermarking, consent flows, limited sharing permissions. But the technology is inherently dual-use. No watermark survives a screen recording. No consent flow prevents a bad actor with access to a target's photos. The same Gemini Omni model that makes editing intuitive also makes forgery fluent.
The real problem isn't Google, or Vids, or this specific update. It's that the industry is collectively shipping AI avatar features into a regulatory vacuum. The US has no federal deepfake law. The EU AI Act covers some use cases but enforcement is fragmented. Every company racing to match Synthesia, beat Sora, or impress Wall Street is incentivised to ship first and ask forgiveness later.
Why It Matters
AI-generated video of real people is about to become as easy as typing a sentence. That's extraordinary for productivity — and catastrophic for trust. The same technology that lets a manager record a weekly update in 30 seconds also lets a scammer impersonate that manager in 30 seconds. We are entering an era where video evidence — historically the gold standard of proof — becomes as unreliable as text.
The companies building these tools have a choice: embed identity verification at the point of generation, or wait for a crisis to force regulation. History suggests they'll wait. The question is how many deepfake-enabled frauds, disinformation campaigns, and reputational attacks will happen in the meantime.