What the report says
The Nexus Project, a research group focused on antisemitism, published findings from tests run on newly created Instagram accounts. According to the report, antisemitic content is both widespread on the platform and largely AI-generated. AI makes such material fast and easy to produce at scale, while reduced moderation lets the flood through.
The algorithm is involved too
Per the Nexus Project, antisemitic content is not merely present but promoted by Meta's recommendation algorithms: engaging with it leads users to see more of it. In one test, the first result for the term "synagogue" on a new account was a video with over a million views; within under an hour, nearly the entire homepage was antisemitic content, more than half of it from verified accounts.
Moderation rollback as an inflection point
The report points to Meta's January 2025 decision to end its partnership with third-party moderators and move to a user-driven model as a turning point. The company framed the change as promoting "more speech and fewer mistakes," arguing its moderation had gone "too far." According to the report, the shift placed the burden of spotting hate speech and misinformation largely on users.



