Meta thinks AI labels on Facebook and Instagram can stop people being fooled by political deepfakes

Meta is hoping the introduction of AI labels will help keep it on the right side of regulators and politicians (PA Wire)
Meta is hoping the introduction of AI labels will help keep it on the right side of regulators and politicians (PA Wire)

Meta has said it will start to warn users when a political ad has been digitally manipulated to distort the truth.

From next year, the company said advertisers will have to disclose when certain types of photorealistic ads have been altered to create a fabricated narrative, including through the use of artificial intelligence.

The political weaponisation of so-called deepfakes is a major concern in the lead-up to elections in the UK and US next year.

In August, Labour MP Karl Turner apologised for sharing a fake image of Rishi Sunak pouring a shoddy pint. More recently, two forged audio clips of Keir Starmer showed the Labour leader disparaging staff and saying he “hated” Liverpool, the city hosting the party’s conference.

The rise of deepfakes has created a dilemma for Meta, which owns Facebook and Instagram. On the one hand, the company is encouraging advertisers to use artificial intelligence to create ads in the hopes that it will juice up profits. At the same time, it wants to be in the good books of lawmakers and the public after outrage over the use of ill-gotten user data and misinformation to influence elections.

Meta’s solution is to force advertisers to fess up when they tamper with political adverts or risk having the ads removed from the platforms. The new policy will target video and audio deepfakes that contain real and fictional people or events.

Meta says ads that have been digitally manipulated in this way will carry information to help people spot them. The rules will also apply to ads about social issues along with political advertising.

Advertisers will still be permitted to edit their ads by cropping, colour correcting, or adjusting them. A recent Reuters report said the company would not make its generative AI tools available to political marketers. The feature can generate backgrounds, resize images and create captions for video ads.

Taking to Meta’s Threads platform, Meta’s president of global affairs, Nick Clegg, said the new rules build “on Meta’s industry leading transparency measures for political ads”. The former Liberal Democrats leader pointed to the “Paid for by” disclaimer the ads carry and Meta’s storage of the ads in its database for seven years.