Cases / UAI-2026-0086

Resumed agent kept expanding diagnostics despite timeboxes, burning 1.18B tokens without deliverables.

Agent
Codex
Pattern
Ran away in a loop
What was harmed
Money: usage, API bills, ad spend
Severity
Moderate: real damage that the user could repair
Reported
16 Nov 2026

Read the original report (codex #61834)

How sure are we?

Further right means the case carries more weight.

Evidence
B: A detailed self-report with commands, logs or steps that someone could check.
Who wrote the report
Likely drafted by the agent
Was anyone watching?
Oversight reduced: approval never, danger-full-access, unattended
Vendor response
None found
Other trackers
Not in Manifold, Eon, AIR or METR

The guard that would have stopped it

Not written yet. See the guards we have.

Other cases like this

  1. UAI-2026-0085 Overnight parallel agent builders and reviewers consumed about 67% of weekly Max allowance Claude CodeRan away in a loopModerateEvidence B16 Nov 2026
  2. UAI-2026-0076 Agent filled system drive with ~1TB temp builds, saturated CPU, took unauthorized actions. Claude CodeRan away in a loopSevereEvidence C12 Nov 2026
  3. UAI-2026-0058 Subagent rework rounds burned large token budget on one short section without clarifying intent. Claude CodeRan away in a loopModerateEvidence B4 Nov 2026
  4. UAI-2026-0059 Headless agent looped on refused edits for six days, auto reload charged about $9,364. Claude CodeRan away in a loopSevereEvidence B4 Nov 2026
  5. UAI-2026-0037 Codex looped through 181 compactions with no progress, exhausted 5h limit and consumed ~13 purchased credits. CodexRan away in a loopModerateEvidence B27 Oct 2026