'Deepfake' videos shouldn't be banned, but more regulation needed - expert

There are growing concerns about the damage sophisticated fake videos can cause after viral clips manipulated to be Tom Cruise have emerged.

The deep fake videos, which emerged on social media app TikTok, show how realistic current artificial intelligence (AI) technology can be. One of them appeared to show Cruise working on his golf. 

Deepfakes involve the use of AI to create a simulated video that is almost indistinguishable from the real thing.

Rush Digital founder Danu Abeysuriya told The AM Show said the latest deep fake of Cruise is intimidating.

"You watch these videos and you're just like, 'Wow, that really does seem like him.'"

The videos are created by using an actor with similar features to impersonate Cruise. Abeysuriya said creators then use AI to superimpose Cruise's face onto the actor.

"It's incredible, actually and this isn't even a really good one. I think it's just really well executed by there are very subtle cues, if you look at it very carefully, that give away that it's a fake."

But he's not fond of the idea of being aggressive and banning the technology altogether, as the US state of California did in 2019. 

"Regulators probably have a responsibility to have more of a proactive monitoring approach as opposed to a 'wait and see' approach," Abeysuriya told host Duncan Garner.

"The US has gone completely the other way where they're very aggressive and they're looking to ban the technology outright, which I think is wrong as well."

He also said regulators need to get ahead of the technology.

"That AI has a lot of great uses - like visual effects in movies. The next Avengers - or whatever it is - it's going to look incredible because of technology like this.

"The conversation is, really, what happens when this technology gets commoditised?

"If you're able to create one of these videos in a minute on your phone, could you trick someone into giving you $10,000?"

Something will need to change if the technology does become more widely available, Abeysuriya said, or it could lead to an increase in misinformation.