One of the areas we’re exploring at Automattic is how generative AI might help us improve video content — for support, training, and product demos.
And one of the most fascinating frontiers is AI avatars: realistic video presenters created entirely by machines, sometimes based on real people.
I wrote about this recently in an article, where I argued that once AI video becomes genuinely usable — fast, high quality, and accessible to anyone — the way we communicate online will fundamentally change.
We’re not quite there yet, but we’re getting close. And before this shift becomes mainstream, I think we need to start asking some hard questions.
Who Owns a Digital Double?
If an avatar is built using someone’s likeness, voice, or mannerisms, who owns it? The individual? The company? The AI provider?
Right now, the legal landscape is vague — and that vagueness could lead to uncomfortable grey areas.
Could a former employee’s AI avatar keep working years after they’ve left?
Should it?
What Does Informed Consent Look Like?
Most consent today is static — a one-time agreement.
But AI-generated likenesses can be reused, remixed, and repurposed indefinitely.
Should consent expire? Be revocable?
What happens when the technology evolves — and a previously cartoonish avatar becomes photorealistic?
Should AI Avatars Be Labelled?
When a customer watches a support video, should they be told if the presenter isn’t real?
Some might not care.
Others might feel misled.
As AI-generated humans become more convincing, the pressure for transparency will only grow. But how should we signal it — subtly, clearly, legally?
What’s the Line Between a Deepfake and a Useful Fake?
The same tech that powers helpful avatars can also create harmful deepfakes.
That means the line between legitimate and malicious use isn’t technical — it’s ethical.
What guardrails should we put in place?
Can content be watermarked or verified without ruining the experience?
A Moment to Choose Well
We’re on the edge of a new visual era. The tools are powerful, and the creative possibilities are extraordinary.
But we’re not just building better videos — we’re helping define what’s acceptable, what’s fair, and what’s trustworthy.
So I’d love to hear your thoughts:
What rights should people have over their AI likenesses? How should companies approach consent, transparency, and long-term responsibility?
At Automattic, we don’t have all the answers. But asking the right questions — now, before the future arrives — feels like the right place to start.
Leave a Reply