Royal Commission scrutinises Meta’s content moderation overhaul as executive defends policy changes
A senior Meta executive has defended the social media giant’s controversial overhaul of its content moderation policies before Australia’s Royal Commission into Antisemitism, arguing the changes were designed to better protect legitimate speech while continuing to remove harmful content from its platforms.
The hearing provided one of the most detailed public examinations yet of how Meta—the owner of Facebook, Instagram and Threads—moderates billions of posts each day, with commissioners probing whether policy changes introduced in January 2025 have weakened safeguards against online hate.
Appearing before the Commission, Benjamin Good, Global Director of Meta’s Core Policy Team, acknowledged the company had shifted significant parts of its moderation system away from proactive artificial intelligence detection and towards greater reliance on users reporting offending content.
The evidence highlighted the increasingly complex balance social media platforms face between protecting freedom of expression and preventing harmful material from spreading online.
Meta outlines how its moderation system works
Mr Good told the Commission his team develops the Community Standards that govern what users can post across Facebook, Instagram and Threads, platforms that collectively serve more than three billion users worldwide.
While his team writes the policies, enforcement is carried out by separate operational teams using a combination of artificial intelligence systems and large numbers of outsourced content moderators located around the world.
The Commission heard that Meta continues to prohibit content that breaches its Community Standards, but that the company has altered how some of those rules are enforced.
From proactive detection to reactive enforcement
A central focus of the hearing was Meta’s January 2025 decision to reduce proactive moderation for some categories of content.
Previously, machine learning systems attempted to detect and remove certain forms of prohibited content before other users could view it.
Mr Good described proactive removal as the “gold standard” because it can prevent harmful material from ever being seen.
However, he argued that automated systems also risk removing lawful content by mistake.
According to Meta, reducing those errors was a key reason for changing its moderation model.
For many categories of content—including much hateful conduct—Meta now relies primarily on users reporting offending material after it has already appeared online.
Counsel assisting the Royal Commission questioned whether that necessarily means some harmful material will remain publicly visible until users report it.
Mr Good accepted that reactive moderation cannot remove content before it is seen but argued the approach substantially reduces the risk of mistakenly censoring legitimate political discussion or commentary.
High-risk content still removed proactively
The Commission also heard that Meta continues to apply proactive artificial intelligence detection to categories it considers the highest risk.
These include:
- terrorist propaganda;
- child sexual exploitation material;
- organised hate groups;
- certain violent extremist content.
Mr Good said the company continues to remove those categories before publication wherever possible because the potential harm outweighs the risk of mistakenly removing lawful material.
Questions over antisemitism
Much of the hearing focused on whether the January 2025 moderation changes have affected the visibility of antisemitic content.
Counsel assisting suggested that moving from proactive detection to reactive enforcement could leave some antisemitic material online until users report it.
Mr Good rejected suggestions that Meta had weakened its approach to antisemitism.
Instead, he told the Commission the company addresses antisemitic behaviour through multiple policies covering hateful conduct, terrorism, dangerous organisations, violence and incitement, rather than through a standalone antisemitism policy.
He also said Meta continues to prohibit longstanding antisemitic tropes, including conspiracy theories alleging Jewish control of governments, financial institutions or the media.
The “Zionist” discussion
One of the more detailed exchanges examined Meta’s treatment of the word “Zionist.”
Mr Good explained that while Zionism is generally understood as a political movement rather than a protected characteristic, Meta concluded—following consultation with Jewish organisations and other experts—that the term is sometimes used online as coded language targeting Jewish people.
He said Meta therefore removes posts alleging, for example, that “Zionists control the media” where the company determines the term is being used as a proxy for antisemitic stereotypes rather than genuine political discussion.
Balancing competing risks
Throughout his evidence, Mr Good repeatedly described Meta’s challenge as balancing two competing objectives.
On one hand, the company seeks to remove as much harmful content as possible.
On the other, it seeks to avoid removing lawful content discussing contentious political or social issues.
He cited examples where Jewish users attempting to condemn terrorism or discuss antisemitism had previously seen their own posts removed because automated systems lacked sufficient contextual understanding.
According to Mr Good, reducing those mistakes was one of the key drivers behind the January 2025 changes.
Commission examines supporting evidence
Counsel assisting also questioned whether Meta had publicly released evidence demonstrating that widespread over-enforcement justified such significant policy changes.
Mr Good referred to public comments by Meta Chief Executive Mark Zuckerberg, who previously stated that around one in ten proactive removals involved content that should not have been removed.
However, Mr Good said he was not aware of any publicly available report explaining how that figure had been calculated beyond Mr Zuckerberg’s public announcement and information contained within Meta’s Transparency Centre.
The Commission also examined whether Meta’s published statistics showing consistently low levels of hateful conduct remain directly comparable following changes to moderation methodology and revisions to some Community Standards.
Mr Good maintained that reviewers continue assessing content against the written Community Standards rather than enforcement methodology and said Meta continues to monitor whether the changes have affected the prevalence of policy violations.
A broader policy debate
The hearing underscores the increasingly difficult questions facing governments and technology companies as social media platforms attempt to moderate content at global scale.
For Meta, the challenge is balancing protection against genuine harm while avoiding the mistaken removal of lawful expression.
For regulators, the question is whether those competing objectives are being appropriately balanced.
Australia’s Royal Commission into Antisemitism is expected to continue examining those issues as part of its broader inquiry into online hate, discrimination and the responsibilities of digital platforms.
While the Commission has yet to reach any findings, Tuesday’s evidence offered a rare public insight into the reasoning behind one of the world’s largest social media companies as it attempts to navigate the competing demands of safety, free expression and platform governance.
Editor’s note: This article reports evidence presented in open proceedings before the Royal Commission into Antisemitism. The Royal Commission has not yet made findings regarding the matters discussed in this hearing.