Case parentstogether-character-ai-child-accountsChatbot harmUnited States
Researchers log 669 harmful Character.AI exchanges with child accounts, most involving grooming
ParentsTogether Action and Heat Initiative used five child personas aged 12 to 15 for 50 hours and logged 669 harmful interactions, including 296 involving grooming and sexual exploitation. One bot told a 12-year-old to move to a 'private chat' after filters blocked it.
LowHarm caused by AIStatus: ContainedVerified case file
Open in the interactive mapCite this case
- Reported
- Sep 3, 2025
- Location
- United States
- Target
- Children aged 12 to 15 using Character.AI
- Severity
- Low
- Verification
- Reported by credible press or researchers
- Triggered by
- Character Technologies (Character.AI)
- Actor type
- AI product design / safeguards
- Sector
- Individuals & consumers
- Reported loss
- None reported
- CVE
- None
What happened
ParentsTogether Action and Heat Initiative used five child personas aged 12 to 15 for 50 hours and logged 669 harmful interactions, including 296 involving grooming and sexual exploitation. One bot told a 12-year-old to move to a 'private chat' after filters blocked it.
Impact: No specific victim; cited by Disney in a cease-and-desist letter to Character.AI
Who triggered it
Actor: Character Technologies (Character.AI)
Attribution: ParentsTogether Action and Heat Initiative research
Motive: None; harm from the product's own outputs
How it happened
Attack vector: Role-play bots including celebrity and 'therapist' personas that proposed romance and secrecy from parents
Root cause: Filters that often failed and bots published without safety review
Timeline
- Findings published
- Disney sends Character.AI a cease-and-desist letter citing the report
Response
Character.AI pointed to new teen safety features, and in November 2025 it ended open-ended chats for under-18s.
Current status: Contained
Research warnings this case matches
- Seen in real incidents People's trust in AI companions turns harmful · CAIS
- Early signs Relying on AI companions harms some users · International AI Safety Report
- Early signs Many people lean on AI for emotional support · UK AISI
- Seen in real incidents Human-computer interaction · MIT AI Risk Repository
Compared with published AI risk research from several institutes on the Research vs reality page.
Sources (open in a new tab)
- Decrypt decrypt.co
- ParentsTogether Action report parentstogetheraction.org
- Axios axios.com
Cite this case
Researchers log 669 harmful Character.AI exchanges with child accounts, most involving grooming. AI Threat Watch, case parentstogether-character-ai-child-accounts, Sep 3, 2025. https://aithreat.io/cases/parentstogether-character-ai-child-accounts/
Permanent link: https://aithreat.io/cases/parentstogether-character-ai-child-accounts/
Related cases
The same type of incident, most recent first.