PLATFORM AUTOMATION · EU

Azar: automation declarations in DSA data.

The monitor turns published statements of reasons from the DSA Transparency Database into comparable monthly values: from detection through the automation level of a decision to the reported restriction.

  • official EU data
  • monthly comparisons
  • platform-level dossiers
  • no AI inference
Source evaluated through · May 26Data series from · Sept 23
169,631published statements in the selected month
17published source months for this platform
19.6%reported with automated detection
68.5%reported with a fully automated decision
THE CRUCIAL BOUNDARY

Automated does not necessarily mean AI.

The DSA data reports whether detection or a decision was automated. It does not name the specific technical method or reliably establish the use of artificial intelligence. This monitor therefore describes automation declarations—not AI use, moderation quality or legal compliance.

automation declaration ≠ AI evidence · volume ≠ quality
AzarMay 26
Detailed flows: latest 3 source months · full history in the open data
THREE SEPARATE DATA STAGES

From detection signal to reported restriction

Choose a platform and month. Connection width follows its published volume. Select a connection for the exact value. Restrictions can overlap and are therefore not added into a misleading total.

non-automated detection → fully automated · 116,148 statements of reasonsautomated detection → partially automated · 32,734 statements of reasonsnon-automated detection → partially automated · 20,288 statements of reasonsautomated detection → not automated · 426 statements of reasonsautomated detection → fully automated · 27 statements of reasonsnon-automated detection → not automated · 8 statements of reasonspartially automated → removed · 32,838 statements of reasonsnot automated → removed · 433 statements of reasonsfully automated → removed · 28 statements of reasonsfully automated → access disabled · 0 statements of reasonsfully automated → demoted · 0 statements of reasonsfully automated → labelled · 0 statements of reasonsfully automated → interaction restricted · 0 statements of reasonsfully automated → age restricted · 0 statements of reasonsfully automated → other restriction · 0 statements of reasonspartially automated → access disabled · 0 statements of reasonspartially automated → demoted · 0 statements of reasonspartially automated → labelled · 0 statements of reasonspartially automated → interaction restricted · 0 statements of reasonspartially automated → age restricted · 0 statements of reasonspartially automated → other restriction · 0 statements of reasonsnot automated → access disabled · 0 statements of reasonsnot automated → demoted · 0 statements of reasonsnot automated → labelled · 0 statements of reasonsnot automated → interaction restricted · 0 statements of reasonsnot automated → age restricted · 0 statements of reasonsnot automated → other restriction · 0 statements of reasonsautomated detection33,187non-automated detection136,444fully automated116,175partially automated53,022not automated434removed33,299access disabled0demoted0labelled0interaction restricted0age restricted0other restriction0

non-automated detection → fully automated116,14885.1% of the relevant stage value

Restrictions can overlap. The right-hand stage therefore uses its own scale and is not an additive continuation of the left-hand stages.

Exact values as a table
Detection × decision
Detectionfully automatedpartially automatednot automated
automated detection2732,734426
non-automated detection116,14820,2888
Decision × restriction
Decisionremovedaccess disableddemotedlabelledinteraction restrictedage restrictedother restriction
fully automated28000000
partially automated32,838000000
not automated433000000

Clearly explained

Does “automated” prove that a platform used AI?
No. The DSA Transparency Database contains automation declarations, but no reliable technical description of the method used. The monitor therefore deliberately refers to platform automation, not AI moderation.
Do statement counts equal the number of moderated items?
No. The unit is published statements of reasons. It is neither a reliable count of unique content nor a count of affected accounts or people.
Why can reported restrictions add up to more than the total?
A statement can mark several restrictions. Removal, disabling, demotion or interaction limits must therefore not be added to infer a new total.
Can I filter German platform values?
Not from this simplified monthly aggregate. It contains no territorial dimension. The monitor therefore reports EU-wide publications by platform and does not invent German subsets.