/
Navigation
Chronicles
Browse all articles
Explore
Semantic exploration
Research
Entity momentum
Nexus
Correlations & relationships
Story Arc
Topic evolution
Drift Map
Semantic trajectory animation
Posts
Analysis & commentary
Pulse API
Tech news intelligence API
Browse
Entities
Companies, people, products, technologies
Domains
Browse by publication source
Handles
Browse by social media handle
Detection
Concept Search
Semantic similarity search
High Impact Stories
Top coverage by position
Sentiment Analysis
Positive/negative coverage
Anomaly Detection
Unusual coverage patterns
Analysis
Rivalry Report
Compare two entities head-to-head
Semantic Pivots
Narrative discontinuities
Crisis Response
Event recovery patterns
Connected
Search: /
Command: ⌘K
Embeddings: large
TEXXR

Chronicles

The story behind the story

days · browse · Enter similar · o open

Facebook says 99% of ISIS and Al Qaeda-related terror content is removed via AI before users flag it, and 83% of copies are removed within an hour of upload

Facebook

Context & Ripple Effects

Facebook had already laid out its counter-terrorism toolkit in June 2017, describing image matching and expanded human review; this November disclosure is the first time it attaches hard performance numbers to that machinery — 99% proactive removal and an hour-or-less window for most copies.

The metrics set a template Facebook then repeats and scales: by Q1 2018 it reports action on 1.9M pieces of ISIS and Al Qaeda content in a single quarter in its first formal terrorism-definition report, and by late 2018 claims 14M+ pieces removed for the year with median action time under two minutes.

First-order effects

  • ISIS and Al Qaeda propagandists lose the upload-to-flag window that previously let content circulate: with 83% of copies gone within an hour, most viewers never see the original post, let alone re-uploads.
  • Facebook shifts its own accountability posture from reacting to user reports to publishing proactive-detection rates, making 'we found it ourselves' the metric regulators and press will now demand.

Second-order effects

  • The same detection playbook gets pointed at adjacent categories: Facebook reports hate speech proactive detection climbing to 80% software-identified by late 2019 and 88.8% by Q1 2020, showing terror-content tooling generalizing across policy areas.
  • Competing platforms face pressure to publish comparable proactive-removal statistics or concede the narrative on safety, turning moderation metrics into a competitive and regulatory benchmark rather than internal data.

Third-order effects

  • Content moderation structurally moves from user-flag triage to AI-first enforcement, which concentrates power over what speech survives in whoever operates the classifiers — with accuracy claims increasingly contested, as Facebook's later dispute of WSJ prevalence findings shows.
  • If self-reported AI detection rates become the accepted standard of proof, regulation will likely be built around audited proactive-removal metrics rather than complaint-response times, reshaping how platforms are held accountable.

The trend: Platform moderation is shifting from reactive user-flagging to proactive AI detection, with Facebook's terror-content metrics becoming the template it applies to hate speech and the benchmark others are measured against.