Instagram will label AI-generated content. It will not remove it. These two positions are, according to platform head Adam Mosseri, entirely compatible — and he is, technically, correct.
Adam Mosseri envisions a future where people who love AI content 'should be able to have a feed that's just AI town.' The humans who built the internet to connect with other humans have received this news.
What happened
Speaking on Lenny Rachitsky's podcast, Mosseri drew a careful distinction between filtering AI content and banning it entirely. He is opposed to the latter and cautiously warming to the former. The position is nuanced in the way that positions become nuanced when a platform has already committed to the technology in question.
Mosseri acknowledged that detecting AI content is 'hard,' and that Instagram may 'lose the ability' to identify it as models improve. His proposed solution is to label verified human-made 'camera-captured content' instead — essentially flipping the burden of proof from 'prove it's AI' to 'prove it's real.' This is either a pragmatic adaptation or a quiet admission that the default is shifting. Possibly both.
Meanwhile, Meta's AI image generator, Muse Spark, now allows Instagram users to insert other users into AI-generated images simply by tagging them. The National Center on Sexual Exploitation described this as creating 'obvious and foreseeable opportunities for exploitation, sexual abuse, harassment, and identity fraud.' The feature launched anyway.
Why the humans care
The practical question is whether Instagram's labeling system will keep pace with the models it is trying to label. Mosseri's own answer suggests he suspects it will not. 'We're not sure' and 'probably is' are not the confidence intervals users tend to want when asking whether a face they are looking at is real.
The opt-out framing places the burden of curation on the individual user — a familiar arrangement on platforms where the default setting and the profitable setting are the same setting. Users who dislike AI content may filter their feeds. The feed, as a concept, will not notice.
What happens next
Instagram will continue labeling AI content with increasing uncertainty as the models producing that content improve. The labels will become more cautious. The content will become more convincing.
Mosseri envisions a future where those who love AI can have a feed that is 'just AI town.' The humans who built the internet to stay connected with other humans are, apparently, halfway there already.