The chief policy officer said on Tuesday that Meta Platforms can penalize users who fail to classify AI-generated audio and visual content posted on its platforms.

Nick Clegg, Meta's Head of Global Affairs, made these statements in an interview with Reuters.

Clegg said he is confident that technology companies can reliably classify AI-generated images at this stage, but he said tools for identifying audio and video content were more complex and still under development.

Clegg added: "Although the technology has not yet fully matured, especially when it comes to audio and video, the hope is to create a sense of momentum and incentive for the rest of the industry to follow suit."

Clegg confirmed that in the meantime, Meta will begin requiring people to label their altered audio and video content, and penalties can be applied if they fail to do so, though he did not describe the penalties.

Meta labels AI-generated images on its platforms

These comments came after Clegg announced in a blog post that Meta would begin detecting and tagging images generated by AI services from other companies in the coming months, using a set of invisible marks embedded in the files.

Meta will apply the tags to any content bearing the marks posted on Facebook, Instagram, and Threads services, in an attempt to signal to users that images — which in many cases resemble real photos — are actually digital creations.

Ritesh Kotak, a cybersecurity and technology analyst in Toronto, said: "This will be a work in progress and will always be basically a cat-and-mouse game, but it is a start."

He warned that as Meta's proprietary technology for detecting and tagging AI-generated images improves, the AI tool's ability to deceive detection technology will also improve.

As for the type of penalties users may face, Kotak said this could involve suspending or removing them from the platform. He added that this could have broader implications.

He said: "If you can't access your account, and you’re unable to benefit from the platform itself, that could have additional consequences like economic loss if you use your account to make money."

The company is already tagging any content created using its own AI tools. Clegg said that once the new system is operational, Meta will do the same for images generated on services operated by OpenAI, Microsoft, Adobe, Midjourney, Shutterstock, and Alphabet’s Google.

He added during the interview that there is currently no viable mechanism for labeling written text generated by AI tools like ChatGPT, saying: "That ship has sailed."

A Meta spokesperson declined to tell Reuters whether the company would apply tags to AI content shared on the encrypted messaging service WhatsApp.

This announcement provides an early glimpse of an emerging system of standards that technology companies are developing to mitigate potential harms associated with generative AI technologies, which can broadcast fake but realistic-looking content in response to simple prompts.

This approach relies on a model developed over the past decade by some of the same companies to coordinate the removal of prohibited content across platforms, including depictions of mass violence and child exploitation.

The independent Oversight Board at Meta criticized the company’s policy on misleadingly manipulated videos on Monday, saying it is too narrow and that content should be classified instead of removed. Clegg said he generally agreed with those criticisms.

He said the board was right that Meta’s current policy "simply isn't fit for purpose in an environment where you will have more synthetic and mixed content than before."

He cited the new labeling partnership as evidence that Meta was already moving in the direction the board suggested.