Meta's global director of core policy tells Australia's antisemitism royal commission it relies on user reports for hate speech takedowns and doesn't track anti Jewish speech separately.
Meta's global director of core policy appeared before Australia's antisemitism royal commission in Sydney on Monday and conceded that the company cannot say whether antisemitic hatred has grown on Facebook or Instagram. Asked directly whether anti-Jewish content had increased on its platforms, Benjamin Good said Meta had no answer, according to the ABC, because the company does not track antisemitic speech as its own category of hate content.
Good defended the company's reactive moderation model on the record. Under it, content stays up until a user reports it. Once flagged, Meta's review teams act on it. The company's own January 2025 "More Speech and Fewer Mistakes" announcement is the documented basis for that posture, and it is the model Meta is now defending at the Royal Commission into Antisemitism and Social Cohesion.
A platform that does not separate antisemitism from other categories of reported speech cannot, by construction, give the inquiry a before-and-after figure on how much anti-Jewish content its users are seeing. That structural gap is what the commission was probing. Counsel assisting the commission pressed Good on whether Meta held any internal data that would answer the question. He said the company had not produced one, according to the Sydney Morning Herald's live coverage of the hearing.
Phrases such as "immigrants are scum", "trans people don't exist", and "gay people are sinners" remained in place during the period under examination, the Australian Financial Review reports, because Meta's review teams concluded they did not violate its current rules. Good argued at the hearing that a reactive model is a more accurate way to keep speech online than preemptive deletion.
Separately, Meta vowed at the hearing to act on coded words such as "Zionism" when used in antisemitic content, The Nightly reported. Meta's reviewers, the company said, can identify the category once it is flagged. The pledge sits alongside the broader reactive posture as an additional removal lane for one category of coded language, not a separate measurement regime.
The testimony now on the record gives an operational definition of "reactive removal" for a platform serving billions of users: a system designed to act on user reports, not to enumerate the categories of hate those reports contain. The commission's opening hearing transcript had set the inquiry's scope as mapping the mechanisms by which antisemitism moves through Australian public life. Meta's appearance is now part of that record.
The accountability question that survives the hearing is also a practical one for any reader. Under a reactive-removal model, content stays up until someone reports it. Without a separate count of anti-Jewish speech among those reports, Meta cannot produce a number from inside. The royal commission's public hearings continue in Sydney this week.