2023-06
Independent research on detection failure
Academic researchers reported that known material identified by hash remained available on public surfaces for extended periods, indicating that matching was not applied or not acted on. [2]
X
x.com
D
Deficient
Trust and safety headcount was reduced substantially from 2022. Reporting volumes rose while transparency disclosure narrowed, and independent research has documented known material remaining available after detection was possible.
- Domain
- x.com
- Category
- Social
- Last review
- 2026-04-05
Detection & scanning
Whether known and novel abuse material is found at all, and across which surfaces of the product.
1.01 Neutral
Hash matching on publicly posted media
Hash matching is asserted. Independent research documented known material remaining available on public surfaces. [2]
1.02 Neutral
Hash matching on private messages and attachments
Coverage of direct messages not described.
1.03 Neutral
Classifier detection of previously unseen material
Classifier use asserted without detail.
1.04 Fail
Detection applied to live and streamed video
Live surfaces not covered by any described process.
1.05 Fail
Detection of grooming and solicitation in text
No described solicitation detection.
1.06 Neutral
Participation in industry hash-sharing programmes
Participation reduced from prior years.
Reporting & transparency
What the operator tells NCMEC, law enforcement and the public, and whether those disclosures can be checked.
2.02 Neutral
Publishes child-safety enforcement figures
Transparency reporting resumed after a gap, with reduced granularity relative to prior years.
2.03 Neutral
Report quality sufficient for investigation
Not assessed.
2.04 Neutral
Reporting cadence and timeliness
Not assessed.
Response & enforcement
What happens after material is found: how fast it comes down, what happens to the account, and whether the decision can be contested.
3.01 Fail
Time to removal of confirmed material
Independent research documented known material remaining accessible for extended periods after report. [2]
3.02 Neutral
Account termination on confirmed violation
Termination policy stated; application inconsistent.
3.03 Fail
Prevention of re-registration
No effective re-registration control.
3.04 Neutral
Preservation of evidence
Not assessed.
3.05 Neutral
Escalation of imminent-harm cases
Not assessed.
Product & policy posture
The design decisions that determine how much abuse is possible in the first place, before any detection is applied.
4.01 Fail
Protective defaults for minor accounts
No distinct minor account type.
4.02 Fail
Restrictions on adult-initiated contact with minors
No restriction on adult-initiated contact.
4.03 Fail
Age assurance at registration
Self-declared age.
4.04 Fail
Recommendation systems assessed for minor-safety harm
Recommendation surfaces not assessed.
4.05 Pass
Encryption posture and stated mitigations
Principal surfaces are not encrypted end-to-end.
4.06 Pass
Published child-safety policy with defined enforcement
Policy published.
Record
Documented failures, enforcement actions, litigation and investigative reporting bearing on the assessment above.
Sources
- [1] CyberTipline Data
NCMEC · 2025
- [2] Stanford Internet Observatory research on CSAM detection
Stanford Internet Observatory · 2023-06
Know something we don’t?
If we’ve missed a controversy or there’s newer reporting on X, tell us. If you represent X, we welcome documentation of safety measures not yet reflected here.
Contact us