Twitter tries to crack down on AI images with Community Notes feature

Twitter users can sign up to create Community Notes that flag questionable tweets and images  (Yui Mok / PA)
Twitter users can sign up to create Community Notes that flag questionable tweets and images (Yui Mok / PA)

As it grapples with the sudden eruption of viral images generated by AI, Twitter is turning to its power users for help.

Elon Musk’s platform is expanding Community Notes to images shared in tweets. The feature essentially puts an explanatory note next to an image when a contributor marks it as potentially misleading.

Twitter says the same note will be attached to all other tweets using the same image, and will be applied to other forms of media (including videos and tweets with multiple pics) in the future.

The update comes on the heels of several AI-generated images that gained notoriety on the social-media app. Community Notes could help Twitter to flag those posts, including the Balenciaga pope pic that went viral a few weeks back. The launch also follows the circulation of an AI-generated image of an explosion at the Pentagon in the US.

Those images were shared by verified accounts, some of which were later banned. It seems that Twitter’s move to open up verification to anyone willing to pay, mixed with Musk’s decision to gut its moderation team, is amping up confusion on the site. Earlier on Wednesday, it was revealed that Twitter’s financial value had plummeted to just a third of the $44 billion (£36bn) Musk paid for it last year.

Currently, the new feature is in the test phase, according to Twitter.

“From AI-generated images to manipulated videos, it’s common to come across misleading media,” the official Community Notes account tweeted.

“Today we’re piloting a feature that puts a superpower into contributors’ hands: Notes on Media. Notes attached to an image will automatically appear on recent and future matching images.”

Twitter said that only Community Notes contributors that are deemed to have met a certain threshold will be able to use the new feature. This so-called impact score will be based on the helpfulness of their prior notes.

Twitter also noted that in the early stages of its trial, its matching algorithm may miss a few tweets that include images with notes.

Twitter rolled out Community Notes globally last December, after debuting the feature in 2021 under former CEO Jack Dorsey. Anyone can sign up to become a contributor as long as they meet certain requirements around account authentication. While everyone else can rate the merit of Community Notes as part of the crowdsourced moderation feature.