The top contributor to X's Community Notes is Web3 Antivirus, a startup that uses an automated process to flag crypto scam posts and has posted 52,000 times
I thought someone was gaming X's crowdsourced moderation; instead, they appear to be helping clean up the platform's mess for free. Bluesky: @mantzarlis.com . X: @craigsilverman See also Mediagazer Bluesky: Alexios Mantzarlis / @mantzarlis.com : When I started researching this story, I thought it was going to be about a grifting private actor abusing Community Notes. Instead, the grifter appears to be X, benefiting from the free labor of a specialized contributor. X: Craig Silverman / @craigsilverman : Scoop: The top contributor to X's Community Notes is a security startup that set up an automated process to flag crypto scam tweets. It's posted 52,000 times! Read the remarkable details at Indicator, my new publication w/@Mantzarlis: https://indicator.media/... See also Mediagazer
Context & Ripple Effects
Community Notes has long depended on contributor participation and voting, but earlier coverage found both that its moderation tools were insufficient for volunteers volunteers handling spam and disinformation and that many responded-to false election posts never received a publicly visible note notes that failed to clear the visibility threshold.
Web3 Antivirus's automated crypto-scam workflow shows a specialized startup supplying a large share of that labor, rather than a broadly distributed volunteer base. That matters because the system's apparent coverage can increasingly depend on a narrow set of highly automated contributors.
First-order effects
- Web3 Antivirus becomes the dominant named contributor to X's Community Notes, with 52,000 posts focused on crypto-scam material; X receives that moderation output without the article indicating it pays for it.
- The concentration makes a single startup's detection process materially more important to what scam content gets surfaced for Community Notes review.
Second-order effects
- A high-volume automated contributor can improve coverage in a narrow abuse category, while also making X's anti-scam performance more dependent on that contributor's rules, uptime, and priorities.
- Other moderation systems may face pressure to separate automated proposals from human judgment more explicitly, especially since prior reporting found Community Notes alone falls short of stopping misinformation.
Third-order effects
- If automated submission becomes normal, crowdsourced moderation is likely to evolve into a hybrid model: machines identify candidate claims at scale and people retain the legitimacy-setting role of rating them. Related coverage later described X's plans for AI agents to propose notes while humans rate them.
- The structural trade-off is between scale and resilience: automation can fill volunteer-capacity gaps, but heavy dependence on a small number of operators could turn an ostensibly distributed trust system into outsourced moderation infrastructure.
The trend: Crowdsourced moderation is shifting toward machine-assisted workflows in which specialized automated systems generate the volume and human participants validate trust-sensitive decisions.