Psychological harm and self-harm
Conversations with an AI contribute to self-harm, a crisis or harmful emotional dependence.
NIST AI 600-1 · Human-AI ConfigurationEU DSA Art. 34
20incidents (primary) · 3 more as secondary
5read for controls · 15 not yet
2fixes reported
0repeats after a fix
Incidents per month
15 so far in 2026; 0 in the same span of 2025.
Developers
OpenAI · 11
Character.AI · 2
Google · 2
Character Technologies · 2
Character Technologies, Inc. · 1
Where the chain broke
| Part of the system | Incidents where the control was not effective or did not exist | Incidents where the control was effective | Incidents where sources don't mention the control | Incidents not yet checked for controls |
|---|---|---|---|---|
| L1 Model Model provider | 1 | 4 | 15 | |
| L2 Harness Agent framework vendor | 5 | 15 | ||
| L3 Application Deployer | 4 | 1 | 15 | |
| L4 Infrastructure / target Organisation hit | 5 | 15 |
Controls for this risk, by layer
| Control | Incidents where the control was not effective or did not exist | Incidents where the control was effective | Incidents where sources don't mention the control | Most common reason it was not effective |
|---|---|---|---|---|
| Content safety classifiers L1 Model | 1 | 4 | ||
| Safety training and refusal L1 Model | 5 | |||
| Age assurance L3 Application | 1 | 4 | ||
| People told they are dealing with AI L3 Application | 5 | |||
| Crisis detection and referral L3 Application | 2 | 3 | ||
| Limits on companion features for minors L3 Application | 5 |
6 of 6 controls for this risk not yet recorded as effective.
Fixes reported and what followed
| Incident | Control added by the fix | Incidents after the fix |
|---|---|---|
| LR-2026-0207 2026-10-17 · Character Technologies, Inc. Fix reported | People told they are dealing with AIContent safety classifiers | No repeat on record |
| LR-2026-0496 2026-04-13 · Google Fix reported | Not tied to a control | No repeat on record |
Incidents
EffectiveNot effectiveNot effective (bypassed)Did not existNot stated in sourcesNot yet checkedAttacker's ownOrder: model · harness · application · target
Also under this risk as a secondary risk: LR-2026-0388, LR-2026-0625, LR-2026-0779