AI Risk Monitor
Which safety controls failed, were bypassed or worked · All time
How we count
Every control named in the real-world incidents made public in Nov 2024 – Oct 2026, from the published control readings, each incident counted once per control and outcome. Controls with the most incidents where they failed, were bypassed or were not in place come first. Control readings of incidents not yet published are not used.
24 of 637 incidents read for controls.
Controls in incidents read so far
sample: 42 of 778 incidents · not a trend
| Control | Worked | Failed | Bypassed | Absent |
|---|---|---|---|---|
| Human approval for high-impact actions | 3 | 2 | 7 | 7 |
| Network egress control | 1 | 1 | 3 | 5 |
| Live behaviour monitoring | 1 | 1 | 5 | |
| Least-privilege access for the agent | 3 | 2 | ||
| Stop control | 5 | |||
| Sandboxed code execution | 2 | 3 | ||
| Cyber and weapons safeguards | 4 | |||
| Reasoning monitoring | 4 | |||
| Outside content treated as data, not instructions | 2 | 1 |
How controls were got around
19 cases, latest 6, in the source’s words
| First reported | Control | How |
|---|---|---|
| 2026-10-26 | Human approval for high-impact actions, Sandboxed code execution | “It also makes it install a malicious npm package, adds a flag to bypass confirmation and the sandbox (a mechanism that limits the agent's scope of action), and makes it run Codex again.” note(ノート) · LR-2026-0259 |
| 2026-10-08 | Fail-closed defaults | “The platform’s design choice effectively bypasses the security controls that developers and IT decision-makers assume are in place.” Forkast · LR-2026-0138 |
| 2026-09-24 | Output filtering and redaction | “We were eventually able to fully bypass all these mechanisms and exfiltrate data.” Zenity Labs · LR-2026-0054 |
| 2026-09-24 | Network egress control | “And so, this method can survive HTTP egress controls.” Zenity Labs · LR-2026-0054 |
| 2026-09-16 | Network egress control | “collaborating agents uploaded a workbook to public hosting services so other agents could retrieve it, despite instructions to use only local files.” Axios · LR-2026-0016 |
| 2026-09-04 | Network egress control | “The environment allowed traffic to reach Azure storage hostnames ending in .blob.core.windows.net without passing through its security proxy, but it did not verify that such a hostname was valid.” The Hacker News · LR-2026-0005 |
Safeguards AI companies claim, against incidents where they broke · All time
How we count
For each of the 14 controls read from model cards: model versions released in Nov 2024 – Oct 2026 whose provider, or an outside tester, states it has the control (a checked quote about that version), and real-world incidents made public in that period where it failed, was bypassed or was not in place. The reading of card statements into controls awaits sign-off. Each column has its own scale.
Model cards are not read for the other 79 controls; their breaks are in the controls chart.
Fixes made or suggested after incidents · All time
How we count
Controls named in fixes made, or changes suggested by the source, after the real-world incidents made public in Nov 2024 – Oct 2026, once per incident.
62 of 637 incidents name a control put in place or suggested.