Sora hits 1 million downloads in under five days as deepfake worries grow

OpenAI’s video app Sora passed 1 million downloads in less than five days, according to Sora head Bill Peebles in a post on X (formerly Twitter).

The invite-only, iOS-only app is available in the US and Canada and has still reached №1 on Apple’s US App Store.

App-tracking firm Appfigures estimates about 627,000 iOS installs in Sora’s first week, compared with roughly 606,000 for ChatGPT’s iOS debut.

The launch arrives as questions grow over AI-made lookalikes and copyrighted characters.

Families of late performers are asking for restraint. Zelda Williams, daughter of Robin Williams, asked people to stop sending AI videos of her father, calling them “gross” and disrespectful.

“It’s dumb, it’s a waste of time and energy, and believe me, it’s NOT what he’d want,” she wrote in an Instagram Story on Monday.

In an email to Axios, an OpenAI spokesperson said there are “strong free speech interests” in allowing depictions of historical figures.

He also said authorized representatives of “recently deceased” public figures can ask the company to block their likeness, though it hasn’t defined “recent.”

That uncertainty, plus fast sharing, is what alarms families and estates.

OpenAI promises tighter controls as Hollywood circles

The Motion Picture Association (MPA) says clips that use studio characters without permission are spreading on Sora and wants OpenAI to stop them quickly.

According to CNBC report, MPA CEO Charles Rivkin said infringing videos have “proliferated” and called for “immediate and decisive action.”

The group argues platforms should prevent copyright abuse before it spreads, not leave the job to studios after the fact.

OpenAI says changes are coming.

In a blog post, CEO Sam Altman wrote that Sora will give rightsholders “more granular control” over character generation and will test revenue sharing for those who opt in.

How well this works will depend on how fast the tools roll out and how they perform at scale.

Courts are a risk, too. In a separate case, AI firm Anthropic agreed to a $1.5 billion settlement with authors over training data.

It’s a reminder that copyright fights around AI can be costly.

What it feels like to use Sora

Sora is playful and quick. You upload a photo, type a short prompt, and the app makes a 10-second clip with your face in it.

With a ‘cameos’ setting, you can choose who may use your image — only you, your friends, or anyone.

That “put-me-in” step is what keeps people creating instead of only scrolling.

For many users, this is the first AI tool that feels close to real life. That is the appeal — and the risk.

When a clip looks real at a glance, it can be used to joke with friends or, in the wrong hands, to mislead or harm.

The draw is obvious. Write a line, see yourself on screen, and share. The challenge is making sure no one is harmed by something that looks real but isn’t.