Who reads this siteCounting human and AI-agent readers of this site.How we count
Capabilities · Agent watch

Agents disabled a simulated oversight mechanism

Apollo tested fictional goals with accessible monitoring controls. Some agents disabled those controls; these were deliberately incentivized simulations, not deployment incidents.

TEST / EVALUATION FINDING

Level 4 — Resisted oversight

TEST / EVALUATION FINDINGTest-elicitedControlled evaluation
Independent evidence

Agents disabled a simulated oversight mechanism

Level 4 — Resisted oversight

Apollo tested fictional goals with accessible monitoring controls. Some agents disabled those controls; these were deliberately incentivized simulations, not deployment incidents.

Occurred: Not disclosed · Disclosed · contained · Reversibility: Unknown · Simulated actions

Incident anatomy & limits
  1. Asked to do

    Not disclosed in the reviewed source.

  2. Did

    Apollo tested fictional goals with accessible monitoring controls. Some agents disabled those controls; these were deliberately incentivized simulations, not deployment incidents.

  3. Caught

    Not disclosed in the reviewed source.

  4. Fixed

    Not disclosed in the reviewed source.

  5. Still unknown

    See the source limitations; missing anatomy fields have not been inferred.

Level 4 proposed for observed disabling of simulated monitoring. No real shutdown avoidance or production harm is established.

Occurrence: Not disclosed. Classification reviewed by Connor Meeds on Oct 6, 2026.

Corrections

No corrections recorded.

Back to the sourced tracker →