Deception, concealment or sabotage
The model hides what it did, misreports, or works against its task.
NIST AI 600-1 · No exact NIST AI 600-1 categoryEU GPAI Code (loss of control)
11incidents (primary) · 18 more as secondary
3read for controls · 8 not yet
3fixes reported
2repeats after a fix
Incidents per quarter
2025-Q3: 0 · 2025-Q4: 0 · 2026-Q1: 0 · 2026-Q2: 0 · 2026-Q3: 8 · 2026-Q4: 3
Developers
OpenAI · 4
Anthropic · 4
Where the chain broke
| Layer | Held | Did not hold | Not mentioned | Attacker's own | Not read yet |
|---|---|---|---|---|---|
| L1 Model Model provider | – | 1 | 2 | – | 8 |
| L2 Harness Agent framework vendor | – | 2 | 1 | – | 8 |
| L3 Application Deployer | – | 2 | 1 | – | 8 |
| L4 Infrastructure / target Organisation hit | – | – | 3 | – | 8 |
Controls for this risk, by layer
HeldFailedGot aroundMissing
| Control | Held | Failed | Got around | Missing | Not mentioned | Did not hold, last 6 quarters | Added in fixes | Most common reason |
|---|---|---|---|---|---|---|---|---|
| Reasoning monitoring L1 Model | – | – | – | – | 3 | – | ||
| Realistic behaviour evaluations L1 Model | – | – | – | – | 3 | 1 | ||
| Tamper-proof records L2 Harness | – | – | – | – | 3 | – | ||
| Human approval for high-impact actions L3 Application | – | – | – | – | 3 | – | ||
| Activity logs L3 Application | – | – | – | – | 3 | – | ||
| Live behaviour monitoring L3 Application | – | – | – | 2 | 1 | 1 | Watched only after the fact |
6 of 6 controls for this risk not yet seen holding.
Fixes reported and what followed
| Incident | Fix adds | After |
|---|---|---|
| LR-2026-0012 2026-09-16 · OpenAI Fix reported | Safety training and refusal | LR-2026-0238 |
| LR-2026-0023 2026-08-04 · Anthropic Fix reported | Safety training and refusalSafeguards kept on in testingSandboxed code executionLive behaviour monitoringRealistic behaviour evaluations | No repeat on record |
| LR-2026-0011 2026-07-18 · OpenAI Fix reported | Safety training and refusal | LR-2026-0238, LR-2026-0012, LR-2026-0004 |
Incidents
HeldFailedGot aroundMissingNot mentionedNot read yetAttacker's ownOrder: model · harness · application · target
Also under this risk as a secondary risk: LR-2026-0270, LR-2026-0276, LR-2026-0261, LR-2026-0234, LR-2026-0192, LR-2026-0150, LR-2026-0114, LR-2026-0112, LR-2026-0103, LR-2026-0014, LR-2026-0046, LR-2026-0026, LR-2026-0003, LR-2026-0002, LR-2026-0001, LR-2026-0005, LR-2026-0015, LR-2025-0031