Meta to mark AI, deepfakes to shield election integrity

Major social media platforms will work to combat voting interference during the Australian election as voters face misinformation during the political campaign.

Meta, the parent company of Facebook, Instagram and Threads, will require posts generated by artificial intelligence or digitally manipulated to carry a disclaimer.

It follows concerns realistic deepfakes - digitally altered images, video or audio that are becoming harder to detect - could be used to spread disinformation and fake news during the election campaign.

Photorealistic content in ads created by Meta AI will carry a label as will generated images users post to Facebook, Instagram and Threads from Google, OpenAI, Microsoft, Adobe, Midjourney and Shutterstock.

More prominent labels will be added to altered images, video or audio "that creates a high risk of deceiving the public on a matter of importance", Meta said.

Ads that don't adhere will be rejected and repeat offenders will face penalties.

Advertisers who run ads about social issues, elections or politics also have to include authorisations including a "paid for by" disclaimer and they're accessible on Meta's publicly available library.

Training sessions are run with candidates and political parties ahead of the election to ensure they have the correct authorisations during the campaign.

Analysing other elections worldwide, Meta found concerns about the rampant spread of deepfakes and misinformation didn't materialise in a significant way.

While such posts were used, they were low in volume, which echoes what Australia's acting electoral commissioner said about deepfakes during the US election.

Fewer than one per cent of fact-checked misinformation on posts about social issues, elections and politics were AI-generated, Meta determined.

The company is also working to track co-ordinated foreign influence operations and says it has taken down more than 200 of these networks since 2017.

"This is a highly adversarial space where deceptive campaigns we take down continue to try to come back and evade detection by us and other platforms," it noted.

Misinformation that could lead to imminent physical harm or election interference will be removed per Meta's policies.

The company is drawing on experiences from global elections, including India, Britain and the US, it said.

Meta continues to work with Agence France-Presse and Australian Associated Press to fact-check posts and those that are debunked will carry a warning label and have distribution limited, so they're less likely to be seen.

It is also working with the Australian Electoral Commission to connect voters to verified information across its platforms, the company said.

Users will receive a reminder to vote on polling day.