Stopping deepfake videos from going viral is like 'whack-a-mole', says electoral commission CEO

Art O’Leary, chief executive of An Coimisiún Toghcháin, told a hearing of the Oireachtas AI committee on truth and democracy that his team spotted a realistic looking video being shared of President Catherine Connolly (pictured) prior to last month’s presidential election and took action with Meta to get the account that shared it suspended.

Stopping deepfake videos made with artificial intelligence from going viral online is like “whack-a-mole" and “hard to get ahead of”, the head of Ireland’s electoral commission has said.

Art O’Leary, chief executive of An Coimisiún Toghcháin, told a hearing of the Oireachtas AI committee on truth and democracy that his team spotted a realistic looking video being shared of President Catherine Connolly prior to last month’s presidential election and took action with Meta to get the account that shared it suspended.

In the fake video, purporting to be a clip from RTÉ News, Ms Connolly was shown saying she was withdrawing from the election.

“We have no regulatory authority here,” Mr O'Leary said. “No investigative ability. We used our special channels with the platform to engage with them. 

The hearing of the Oireachtas AI committee on truth and democracy was told that Ms Connolly was shown saying she was withdrawing from the election in a deepfake video.

"We did so before midnight [after spotting it at 10.30pm]. By the time everyone got out of bed the next morning, the account was suspended.” 

Mr O’Leary doesn't know where the piece of AI-generated content came from, but said it was caught it within minutes of it going online and the “system worked” with Meta. He added that some social media companies “are less willing” to act as Meta did.

Coimisiún na Meán

Meanwhile, the head of Ireland’s media regulator Coimisiún na Meán said social media firms should face similar obligations around plurality that traditional broadcasters do.

"In order to ensure that audiences are exposed to a variety of voices, views, and opinions, broadcasters are obliged to make sure that news and current affairs output is presented in an objective and impartial manner without any expression of the broadcaster’s own political views,” Jeremy Godfrey said.

“There are currently no corresponding obligations that apply to content feeds on social media, or to the output of AI chatbots.” 

He said social media feeds are a primary source of news, particularly for younger people, but these feeds can also offer users “rage bait” and send them down rabbit holes.

Mr Godfrey said a platform “isn’t the editor in the way a broadcaster is but they have a big role in selecting” what goes into a person’s social media feed.

What's your view on this issue?

You can tell us here

When pushed by People Before Profit’s Paul Murphy on why guidelines on such “recommender” systems weren’t included in the regulator’s Online Safety Code, Mr Godfrey said they believed it was best covered by separate legislation.

He acknowledged that recommender systems can have a negative effect on children in particular and, when it’s done to purely keep people scrolling on the platform, “they don’t keep users safe”.

AI 'corrosive to democracy'

Furthermore, Dr Dan McQuillan, a senior lecturer in creative and social computing at Goldsmith at the University of London, hit out at AI as being “corrosive to democracy” and said it is being used to “substitute slop for critical thinking”.

He said governments should be very careful when considering whether to embed AI into its work across departments. 

He added that Grok, the AI tool of social media giant X, had been "directly interfered with by its owner" in the responses it gives. 

“At best, the alleged benefits to healthcare and so on really amount to algorithmic Thatcherism,” Mr McQuillan said. “A more likely outcome is that widespread AI adoption will strengthen the far right.”