Meta in hot water over explicit AI images

Meta’s independent Oversight Board is investigating Meta’s AI content policies and its actions after two explicit AI-generated images were posted to Facebook and Instagram.

How Meta detects and responds to AI content

Meta currently uses a combination of AI and human review to detect explicit AI content on its platforms and has recently announced its plans to add “Made with AI” labels to a wider range of content, to provide more clarity and prevent deep fakes from harming the public.

However, the Oversight Board has been brought in to investigate whether Meta’s enforcement policies surrounding AI-generated deep fakes are effective and also consistent.

"The Board selected these cases to assess whether Meta’s policies and its enforcement practices are effective at addressing explicit AI-generated imagery."

What was the explicit AI content found on Meta’s platforms?

The first incident happened in India, where an Instagram user reported an explicit AI-generated image, of a public figure, to Meta, asking them to take it down.  

Meta failed to do so, and the ticket was automatically closed after 48 hours. The user appealed Meta’s decision, but again, the ticket was closed, and the image remained on Instagram.

The user then reported the image to the Board, and Meta swiftly removed it, citing it had “breached community standards on bullying and harassment”.

In summary: It took two reports to Meta, and a report to the Board for the explicit AI image to be removed.

The second incident happened in the US, where Meta’s systems detected an AI-generated explicit image of a public figure on Facebook, and removed the image immediately.

The difference in how Meta approached the two incidents is concerning–especially as both images were in breach of Meta’s AI content policies–and explains why the Board has stepped in to investigate.

“The Board believes it’s important to explore whether Meta’s policies and enforcement practices are effective at addressing this problem.”

The Board is currently investigating both cases (they have asked for public comments by April 30th), and will post their decision in a few weeks.

In both cases, Meta has now taken down the explicit content, and the Board will not name the individuals involved to protect them from further harassment.

‍