Meta
GLOBAL PUBLIC POLICY & INTEGRITY
EXECUTIVE DIRECTIVE
Integrity Operations & Trust Architecture
TO:
Integrity Operations, Policy Development, Product Engineering & Legal
FROM:
Mark Zuckerberg, CEO & Joel Kaplan, Chief Global Affairs O
ffi
cer
DATE:
September 17, 2026
SUBJECT:
Meta Content Moderation Memo
1. STRATEGIC CONTEXT & PURPOSE
Over past election cycles and global crises, Meta built intricate, multi-layered systems to police user discourse across
Facebook, Instagram, and Threads. While built with the intention of protecting platform safety, overly complex rules
and expansive automated enforcement mechanisms led to signi
fi
cant unintended consequences: millions of regular
speech instances incorrectly suppressed, increased user distrust, and perceptions of institutional bias.
This
Meta Content Moderation Memo
establishes the foundational operational pivot toward rightsizing platform
governance. We are simplifying rules, dismantling legacy centralized third-party fact-checking programs, scaling
crowdsourced community context, and resetting automated detection thresholds to prioritize open expression while
decisively suppressing severe criminal harm.
2. CORE POLICY ARCHITECTURE: SHIFT TO COMMUNITY NOTES
To eliminate centralized arbiters of truth and empower organic community consensus, Meta is transitioning its
integrity framework across three structural pillars:
Sunset of Third-Party Fact-Checking:
The centralized external fact-checking contractor program is formally
phased out and replaced with a distributed, crowdsourced Community Notes framework across Facebook,
Instagram, and Threads.
Cross-Perspective Consensus Engine:
Community Notes require agreement between contributors who historically
disagree in their evaluation patterns, preventing ideological brigading before context notes become visible on user
feeds.
Removal of Top-Down Demotions:
Posts subject to community review will carry contextual annotations rather
than aggressive algorithmic reach penalties or automated shadowbanning.
Foundational Principle:
Complex moderation architectures make systematic mistakes. If our automated
fi
lters
accidentally censor even a small fraction of posts, millions of authentic voices are silenced. We must default to open
expression unless clear, severe, and high-con
fi
dence policy boundaries are breached.
3. HIGH-SEVERITY AUTOMATED ENFORCEMENT THRESHOLDS
Automated algorithmic takedowns are now strictly limited to high-con
fi
dence, severe harm categories. Proactive
automated removals are con
fi
ned to child sexual abuse material (CSAM), violent terrorism, illegal narcotics
tra
ffi
cking, weapon sales, and coordinated
fi
nancial fraud.
Meta Platforms, Inc. Con
fi
dential
—
Internal Operational Policy
Page 1 of 2
