Content Moderation 'Art, Not Science': Live-Stream Security Gaps Exposed

In cybersecurity, a system that relies on 'art, not science' is a system with exploitable gaps. Kick's testimony reveals that live-streaming moderation is a black box strained by false reports and subjective calls — a perfect environment for malicious actors to slip through.

In a dramatic session of Australia’s Royal Commission into Anti-Semitism, social media platforms Meta and Kick laid bare the deep subjectivity and operational gaps that allow hate speech to flourish online. Kick, a live-streaming platform with over 100 million users, had its general counsel unable to definitively state whether calling Jews “evil rats and subhumans” — a classic Nazi propaganda trope — would violate the company’s content guidelines. The official, Tiat Oon Ooi, described hate speech detection as “more art than science,” conceding that while calls to violence are obvious, more insidious slurs exist in a gray area not codified into clear rules. Meta, the $1.5-trillion-plus parent of Facebook, Instagram and Threads, separately told the inquiry that it treats hate speech, including anti-Semitism, differently from “truly heinous” content such as terrorism glorification, implying a tiered enforcement model that critics say creates a tolerance zone for borderline hate.

“

Kick's testimony reveals that live-streaming moderation is a black box strained by false reports and subjective calls — a perfect environment for malicious actors to slip through.

The testimony comes at a critical regulatory moment. Australia is weighing new online safety legislation, the EU’s Digital Services Act is imposing systemic risk assessments, and the UK’s Online Safety Act demands platforms remove illegal content proactively. Meta’s distinction between hate speech and “truly heinous” content suggests that content classified merely as hate may not face the same rigorous takedown pressure, potentially allowing harmful narratives to incubate. This tiered approach, combined with Kick’s admission that its moderation is largely outsourced to a team in Serbia and plagued by a high volume of false reports, reveals a brittle enforcement architecture. Kick’s removal rate for reported posts is far lower than peers, which Ooi attributed to users generating false alarms against disliked streamers, effectively diluting the signal for genuine hate speech.

The implications ripple beyond these two platforms. Advertisers have long fretted over brand safety, pulling spend when ads appear next to extremist content. Meta’s policy categorization and Kick’s nebulous standards signal to media buyers that hate speech remains a stubborn adjacency risk, especially on live-streaming platforms where real-time moderation is inherently harder. For regulators, the “art vs. science” framing undercuts any assurance that platforms can comply with emerging duty-of-care obligations. If a company’s top lawyer cannot definitively classify the most dehumanizing anti-Semitic trope, how can the company credibly assert it’s taking “reasonable” steps to protect users? This may become a pivotal argument in future enforcement actions and class-action lawsuits.