Meta's own safety process got overridden at the top, and it's a bad sign for how the company handles real-world harm on its platform.

• A 40-person team spent six months building a policy for when a world leader's posts cross the line into inciting violence, triggered by India's Modi using language flagged as a warning sign for ethnic violence against Muslims.
• After the team presented its carefully negotiated proposal, Mark Zuckerberg rejected it and wrote his own rule over a weekend: never censor a politician's voice.
• That replacement rule was vague and hard to apply in practice, since it did not even clearly define who counts as a "politician."
• Staff were reportedly demoralized, since a leader overruled months of expert work backed by policy, legal, and human rights specialists inside the company.
• Within two years of this type of decision-making, social media was linked to ethnic violence in Ethiopia that killed large numbers of people.

Outlook: Expect continued scrutiny of Meta's content moderation choices as critics argue top-down overrides keep undercutting internal safety teams.