Victims of AI deepfakes could sue for emotional damages under new bill

Australians who share AI deepfakes of another person without that person's consent could be sued or face steep fines, under a new proposal before federal parliament today.

Independent senator David Pocock, who is introducing the bill, said the government was yet to respond to rapid advances in the field of AI, and urgent action was needed.

"I'm concerned that the government has dropped the ball," Senator Pocock told the ABC.

"Currently, unless a deepfake is sexually explicit, there's very little that you can do as an Australian."

He said Australians lacked "basic safeguards" against having their likeness copied and used for scams, commercial exploitation, disinformation and a range of other purposes.

The proposal seeks to add a dedicated complaints framework to the Online Safety Act, granting the eSafety Commissioner powers to demand deepfake removals and issue immediate fines.

Under the changes, individuals who share a non-consensual deepfake could be fined $165,000 up-front, while companies that fail to comply with a removal notice could pay as much as $825,000.

As well as the initial fine, individuals caught sharing non-consensual deepfakes could also be hit with a second fine for ignoring removal notices.

Additionally, proposed changes to the Privacy Act would allow Australians to bring civil lawsuits and sue perpetrators directly for financial compensation.

Under Senator Pocock's proposal, the victim would not have to prove they had suffered financially in order to bring a case, in recognition of the emotional harm caused by deepfakes.

"As humans — our likeness, our face, our voice, our mannerisms — that's a big part of who we are," he said.

The bill also contained important limits, Senator Pocock said.

It would carve out uses such as journalism, satire, and good faith uses by law enforcement, with similar exemptions and defences to defamation law, and an additional exemption for children.

The bill was also designed with the goal of sparing people who share non-consensual deepfakes without realising it, while still catching out those who do so knowingly or recklessly.

Last year, the crossbencher created AI-generated videos of Prime Minister Anthony Albanese and then-opposition leader Peter Dutton ahead of the election, to highlight how easily deepfakes could distort political debate.

He said the changes proposed in his bill would also apply to political campaigns.

"We have to draw a line in the sand and say, this is not on — you cannot deepfake someone without their consent," he said. 

Crossbench moves on AI as government mulls response

Senator Pocock's bill is not the only effort from the crossbench to tackle AI harms, fuelled by a sense that the government has fallen behind.

In July, independent MP Kate Chaney introduced a private members' bill to criminalise AI apps and tools that generate child sexual abuse material.

"I spoke to the attorney-general's office, and there was an acknowledgement that, yeah, there probably is an issue here. But still, we've seen no action on it," she said.

The government passed a law in August 2024 to criminalise sharing non-consensual sexually explicit deepfakes, although it did not criminalise the act of creating such material.

Ms Chaney and Senator Pocock said there was now an urgent need for far broader protections beyond the issue of sexualised deepfakes.

"The crossbench is stepping up here because the government has been missing in action," Senator Pocock said.

"Talking to Australians, they're concerned about the misuse of AI," he said.

Last week, Ms Chaney raised similar concerns about the lack of protections, calling for Australia to establish its own AI Safety Institute.

"The US, the UK, Canada, Japan, Singapore all have an equivalent. Australia has supported the idea of them but has not yet actually taken action on putting one forward," she said.

The government has been considering its wider policy response to AI after mothballing a draft proposal for specific AI regulation, driven by the previous minister for industry and innovation, Ed Husic.

Labor's rhetoric shifted in the middle of the year to instead focus on the economic promise of AI, as the Productivity Commission called for AI laws to be "a last resort".

In August, new minister Tim Ayres told the ABC the government was not "rushing" towards a regulatory response, and was waiting on the outcome of a review due at the end of 2025.

"I want to make sure that the government makes good decisions here … to deliver the right outcome," he said.

Senator Pocock said the government began consultation on AI two years ago and since then had done "absolutely nothing with it".

"At the same time … swallowing the lines from big AI companies who are promising the world in terms of productivity, but not actually talking about the very real risks," he said.