YouTube review results in smear campaign remaining online

A YouTube video describing a Bondi survivor as a “front man” with a “degree in theatre” is still online after human moderators claimed it did not breach rules.

Jewish Australian Arsen Ostrovsky was the target of both the video and a deepfake smear campaign which began shortly after having been injured in the December 14 terror attack which killed 15 innocent people celebrating Chanukah.

Rachel Lord, a senior manager at YouTube Australia defended the hurtful content being “still online” more than six months later in her testimony at the Royal Commission on anti-Semitism and Social Cohesion.

The video involves four men alleging Mr Ostrovsky was “a front man”, “Zionist” and “intelligence asset” for Israel with a “degree in theatre”.

“This dude is clearly out using it as his own propaganda,” one of the creators said.

“I don’t believe that blood, I mean he definitely dressed that … up for sure.”

Ms Lord said the video had been “thoroughly” reviewed at “quite senior levels” resulting in the content being determined as “non-violent”.

“What I have been told from the teams that have reviewed the video, is that it is non-violative and it remains on the platform,” she told the commission.

“From what’s been shared with me is that part of the policy is looking at denial of the event itself or denial of victimhood, not questioning motivations and actions around it.”

“Zionist is obviously not a protected group under our hate speech policy, but where it is being used … as a proxy for a protected group and that content is violative, the content will come down but the teams have reviewed this thoroughly.”

YouTube’s hate speech policy prohibits material that “denies or minimises a well-documented major or violent event or the victimhood of such an event”, the commission heard.

In response, Ms Lord claims their moderators were “looking for … denial that the event took place or that people died as a result of the event”, which she claimed was not present in the video attacking Mr Ostrovsky.

The YouTube representative agreed if the same video was reported again, “absolutely nothing would happen” and the video would remain online.

“If it were flagged, it wouldn’t be removed because it’s not deemed to be violating.”

Ms Lord said she first became aware of the video five months after the Bondi attack in May and defended their moderation processes as “robust”.

The eSafety Commissioner reportedly contacted YouTube in the “immediate aftermath” of the Bondi attack, who subsequently claimed its staff were “working around the clock to monitor for violating footage”, the commission heard.

“YouTube was focused on ensuring Australians and all users around the world have access to high quality information about the tragic event,” Ms Lord.

Humans “complicate” TikTok moderation

An anti-Semitic TikTok scrubbed after AI moderation was reinstated after an appeal by the creator overseen by a human, the company has said.

Details of the clip cannot be published for legal reasons, but the U-turn shows that human judgement “can sometimes complicate” correct AI decisions, TikTok said.

The backflip was discussed by TikTok’s Global Head of Policy, Trust and Safety Zachary Hecht who travelled from New York during his appearance at the to appear before the Royal Commission on anti-Semitism and Social Cohesion.

The Dor Foundation subsequently reported the material and gave evidence of the incident to the royal commission at an earlier hearing.

After review by at least three different human reviewers, the content was finally removed in early July.

“It does demonstrate that in some instances, automated moderation might be getting the decision correct and then human judgement can sometimes complicate that,” Mr Hecht said.

“It would also be an instance where I would want to ensure that anybody that reviewed this content would have targeted intervention to receive more training on what the correct outcome was in this case.”

“I do have confidence that once this was escalated to more senior members of the team … they did end at the right conclusion.”

TikTok’s superiority undermined

Close to 100 per cent of TikTok’s moderated posts in Australia are proactively removed before being reported by users.

The social media company boasts a “substantially superior performance in terms of removal rates” when compared to other platforms, according to senior counsel Richard Lancaster.

This is despite a lack of data regarding how much harmful content is missed by TikTok’s moderation processes.

Mr Hecht revealed that “98 per cent of content” identified as violating the company’s community guidelines was “proactively removed” before any complaint reports.

“This is something that we disclose in our community guidelines enforcement report which is available publicly,” he said.

“That means that we have removed that content before any user reported it to us.”

Counsel Barrister Nicholas Bender pointed to the fact that TikTok does not maintain data on the content “missed” by their moderation process or report on the amounts of anti-Semitic material being shared online.

Mr Bender shared data which shows that total removals of TikTok posts “violating safety and civility guidelines” in Australia have decreased from 338,000 in 2023 to 67,000 in the first quarter of 2026.

“So all that tells one is that year on year, TikTok is detecting and removing less violating content, doesn’t it?” he asked.

In response, Mr Hecht said there was a “potential that there is less violative content being posted” in the first place.

His assertion contradicted earlier testimony from Dr Oboler which indicated a “sharp rise” in anti-Semitism since the October 7 attack.

Mr Hecht described a process of “dynamic” moderation executed by both automated and human review.

“We’re regularly trying to identify the best way to ensure that bad actors are not engaging on the platform, but also educate users so they might know that they have to change their behaviour to remain on the platform,” he said.

While harmful content is identified and reviewed, users are usually banned after “multiple violations”.

“In addition, when there are absolutely egregious violations, we ban a user after a single instance,” he told the commission.

“It’s not necessarily covered by the hate speech and hateful ideology section but we also prohibit incitement to violence.”

“So when there’s an active incitement to violence, which may be against the Jewish people and would therefore be anti-Semitic, an account would be banned right away.”

At the start of hearing block three, the royal commission heard a “Jill all Kews” post successfully dodged TikTok’s moderation processes.

The content targeted the president and board member of Australia’s largest Jewish school, Judy Lowy, referring to her as “Genocide Judy”.

Another Jewish mother gave testimony about her daughter being the target of anti-Semitic slurs on TikTok.

“I wish Hitler had gassed your family so we wouldn’t have had to meet,” one comment wrote to the Year Seven schoolgirl.