The Royal Commission on Antisemitism and Social Cohesion held its third public hearing block this week, dedicated to investigating the spread of antisemitic content online and in traditional media. It followed two earlier blocks this year that heard firsthand accounts of antisemitism from Jewish Australians and probed the security failures surrounding the December 2025 Bondi attack.
This block examined the nature, prevalence and drivers of antisemitism and other hateful speech on social media platforms and other online forums, as well as how public media organisations identify and respond to antisemitism, the impact of media coverage on the daily lives of Jewish Australians, and the role of the Australian Communications and Media Authority in handling complaints. Commissioner Virginia Bell AC SC also weighed how effectively law enforcement, regulators and platforms cooperate to combat hateful and extremist content, and what further measures might address the gaps identified so far.
Dr Andre Oboler, founder and CEO of the Online Hate Prevention Institute (OHPI), returned to the witness stand, having also given evidence in the Commission’s first hearing block, to walk the Commissioner through thousands of examples of antisemitic content gathered from seven major platforms, and to warn that automated moderation systems are failing to keep pace with a surge that, on several platforms, has now overtaken its original post-October 7 peak.
Oboler, a computer scientist and adjunct associate professor at La Trobe University’s Law School, described OHPI’s monitoring methodology: trained analysts spend dedicated hours “snowballing” through platforms, capturing an initial piece of antisemitic content, screenshotting it, working through the comments beneath it, then following individual users who engaged with the post to see whether they too had shared antisemitic material. Each item is classified into one of 27 sub-categories drawn directly from the International Holocaust Remembrance Alliance’s working definitions of antisemitism and Holocaust denial.
The report tendered to the Commission, which required a bundle of platform statements to be temporarily withheld due to ongoing confidentiality discussions, drew on roughly 500 examples each from X and Meta, and 250 from five other platforms, all of which were reported to the companies at least twice by OHPI staff.
Oboler’s data showed divergent patterns across platforms. On Facebook, he said, levels of antisemitic content spiked sharply after October 7, 2023, then rose again in January and April this year to points “significantly above” the original post-October 7 high, a rise he linked partly to escalation in the Iran conflict. Telegram, by contrast, was dominated almost entirely by “traditional” antisemitism tied to far-right channels, some of which he said were promoting the Iranian Revolutionary Guard Corps and functioning as content repositories for supporters to redistribute.
Platform responsiveness varied sharply, Oboler told the Commission. TikTok had a “relatively good” removal rate once content was reported, but its heavy reliance on public flagging over proactive detection meant much of the material had a short lifespan, disappearing before consequences could follow. YouTube’s multi-layered system, which automatically screens content and queues borderline material for human review before it reaches the public, meant a lot of hate speech never went live in the first place. Meta, he said, had actually pioneered that proactive automated approach and, for years, reported catching over 90 per cent of hate speech before anyone saw it, but the company scaled it back sharply in January last year, resulting in “a massive drop in removals” visible in Meta’s transparency reporting.
Oboler was critical of relying on public reporting as the primary safeguard, arguing it shifts the burden of moderation onto the communities being targeted, who must first be exposed to the content before they can flag it, rather than onto the platforms to prevent harm before it happens.
On X specifically, Oboler was pointed. He told the Commission the platform had become significantly less cooperative since Elon Musk’s takeover, which saw roughly 80 per cent of its trust and safety staff let go and a philosophical shift toward “more speech, fewer limitations.” That change, he said, allowed previously banned users back on and made the environment progressively more toxic. He noted Musk himself had amplified content that “promotes antisemitism and pushes hate,” pointing to the gesture Musk made at a Trump event that was widely described as a Nazi salute.
Describing X’s approach to transparency, Oboler said: “We’ve got a temperature control on the wall, and rather than fixing the temperature, what they’re doing is tampering with the monitoring device,” removing the material being publicly tracked, he said, rather than addressing the underlying problem. He contrasted X’s well-resourced engagement with regulators in Europe against what he described as years of minimal contact from the platform’s own staff in Australia.
He was particularly critical of arguments that antisemitic content directed at Israel or Zionism should be exempted from action because it is “political,” telling Bell: “racism is always political … The first question is not whether it’s political, the first question is whether it’s racist.” He drew a comparison to Nazism and Australia’s former White Australia policy as examples of racism that was also, undeniably, political.
Oboler also challenged the way platforms report their own success rates, drawing on the old description of the internet as the “information superhighway.” He told the Commission platforms today effectively function like separate stretches of that highway, each a virtual monopoly making up its own rules regarding speed limits, licensing, and seat belts. Safety standards, he said, vary wildly between platforms and are generally low across the board.
Applying that analogy to Meta’s own transparency reporting, he pointed to the platform’s stated prevalence rate of antisemitic content: 0.01 per cent of views, or one in 10,000.
“If we were to use that traffic analogy… we’re going to have an incident for every 10,000 metres a person travels,” he told the Commission. “And you consider how many metres a person travels, then multiply that by the population that’s travelling, you’re actually talking a huge volume of incidents, and it’s just the way it’s reported that makes it look low,” he said.
The hearing continues, with the Commission’s fourth block scheduled to move to Melbourne from 13 July.
