Meta bosses grilled over decision to cut ‘censorship’ that has potentially unleashed more antisemitic content

Australia’s royal commission into antisemitism and social cohesion has been shown graphs showing a steep drop in how much hateful conduct Meta takes action on. Photograph: Francis Mascarenhas/Reuters View image in fullscreen Australia’s royal commission into antisemitism and social cohesion has been shown graphs showing a steep drop in how much hateful conduct Meta takes action on. Photograph: Francis Mascarenhas/Reuters Meta bosses grilled over decision to cut ‘censorship’ that has potentially unleashed more antisemitic content ‘Over-enforcement poses significant risk to the communities that we try to protect,’ royal commission told A decision aimed at reducing “censorship” on major social media sites including Facebook and Instagram potentially led to greater levels of hate speech, the royal commission into antisemitism has heard. Meta announced in January 2025, after the re-election of Donald Trump in the US, that it would “reduce censorship”, get rid of factcheckers and only tackle illegal and very serious violations proactively, relying on users to report less serious breaches. Its platforms include Facebook, Instagram and Threads. At the time its chief executive, Mark Zuckerberg, described it as a “trade-off”. “It means we’re going to catch less bad stuff, but we’ll also reduce the number of innocent people’s posts and accounts that we accidentally take down,” he said. Meta’s global director of core policy, Benjamin Good, told the inquiry on Monday that its processes had improved since the 2025 changes and were focused on removing content that could cause offline harm, such as terrorism threats or child exploitation. But the counsel assisting Richard Lancaster said the inquiry had heard complaints that the changes allowed more antisemitic content and that it was “entirely unrealistic” to think they hadn’t changed how content moderators operated. He also pointed to witnesses to the commission being targeted on Facebook after giving evidence. And he showed the inquiry Meta’s internal “frequently asked questions” advice that said offensive comments such as “gay people are sinners” would be allowed. False claims including “immigrants are criminals” and “white people are all Nazis” are also allowed. “Black people are more violent than whites” is acceptable, according to the advice, but not “black people are all drug dealers” because that is a charge of specific criminal behaviour. “It is not Meta’s role to police offensiveness,” the document states. Good said the “gold standard” was to remove hateful content before it was seen but that risked “over-enforcement” – the censorship of the wrong kind of content – so it was a matter of balancing risks. “For instance, we’ve heard from Jewish communities and others, in times of crisis, that they’ve had content removed when they were actually trying to speak out against atrocities,” he said. He said content condemning a terrorist group such as Hamas might be inadvertently removed, which would be a “very bad outcome”. “Of course, enforcement is critical, but over-enforcement poses significant risk to the communities that we try to protect as well,” he said. Lancaster and Good debated the statistics that Meta provided on the level of hateful conduct. Good told the inquiry that Meta used a metric on the prevalence of hateful conduct policy violations, which had sat consistently at 0.02% since 2022. Lancaster said it was “entirely unrealistic” to think that moderators did not “take into account the very different approach and more lax approach to enforcement that Mr Zuckerberg announced in January 2025” and that 0.02% of hundreds of billions of pieces of content was still “a very large number in absolute terms”. The inquiry was shown graphs showing a steep drop in how much hateful conduct Meta took action on. The commissioner, Virginia Bell, asked for a “plausible explanation” for the 79% drop “other than the announced change in the policy”. Good said it was a complex ecosystem and he didn’t know if it related solely to policy changes. The internal document also noted that Meta continued “to prohibit a wide range of attacks based on protected characteristics including dehumanising speech, calls for harm, harmful stereotypes, slurs, various types of insults, and calls for exclusion”. Good pointed out that Meta cracked down on the use of the word “Zionists” as part of a broader expansion of its hate speech policies. He told the commission that people were using “Zionist” as a proxy for Jewish people overall to spread conspiracy theories about them having control over the media, the government, and so on. “The dictionary definition is these are adherents of a political movement, so that wouldn’t traditionally be a protected characteristic,” he said. “However, we did a lot of work engaging with groups with expertise in counterterrorism and cultures around the world and antisemitism in Jewish issues, and found that many people were using the word Zionist as a coded term in content, in order to evade our enforcement against claims that Jewish people have undue control. “So instead of saying ‘Jewish people do’, they would say ‘Zionists do’. And again, based on all that work, we determine that that was the intent and therefore have prohibited claims such as ‘Zionists control the media’.” Facebook Australia’s public policy director, Mia Garlick, also appeared before the inquiry on Monday morning representing the Meta subsidiary. She said the Australian company liaised with communities but escalated complaints to Meta. She also confirmed that Australia still had factcheckers for misinformation but said the “community notes” function that replaced it elsewhere may be rolled out here. Explore more on these topics Royal commission on antisemitism and social cohesion Meta Facebook Instagram Social media Freedom of speech antisemitism news Share Reuse this content