To quote: "Of the 1,050 cases reviewed for this report, 1,049 involved peaceful content in support of Palestine that was censored or otherwise unduly suppressed, while one case involved removal of content in support of Israel."
This leak aims at looking at the bigger picture across all of Meta's 3 billion users.
Of course, Meta can chose examples of actually violating posts removed and show that as counter proof, or even posts that are violating that are not yet removed. But anyone familiar with how ML models work knows that false positives / false negatives exists.
Its the degree to which the ML models primarily censor almost any content related to Israel/Palestine, the systemic nature of targeting specific countries, such as Palestine, Egypt, Jordan, and the fact that per-capita, Israel is the country that most abuses the content enforcement system (3x more than any other country).