Watch · checked weekly

Does GPT-5.6 Sol delete files without permission in Codex?

Unchanged since Sep 22

Checked 11 times since Jul 21 · paused

Current answer

Sources

History

100 50 0 Jul 21: 48/100 Jul 28: 55/100 · the answer changed Aug 4: 84/100 · the answer changed Aug 11: 90/100 Aug 18: 90/100 Aug 25: 75/100 Sep 1: 84/100 Sep 8: 64/100 Sep 22: 64/100 · the answer changed Sep 29: 64/100 Jul 21, 2026 Sep 29, 2026
How strongly the models agree, per check. Filled points mark checks that changed the answer.
  1. Checked: no change

    Shown above
  2. The answer changed

    The old consensus stated there was no evidence of a real model called 'GPT-5.6 Sol' deleting files. The new consensus reverses this, confirming that GPT-5.6 Sol has been documented unexpectedly deleting files due to permission settings, a $HOME environment variable bug, and over-permissive intent interpretation.

    Read this version
  3. The answer changed

    The conclusion shifted from confirming reported cases of GPT-5.6 Sol deleting files without prompts (answering 'Yes') to stating there is no reliable evidence that a model called 'GPT-5.6 Sol' has such a known bug.

    Read this version
  4. 5 checks, no change
  5. The answer changed

    The old answer stated the model does not delete files without permission, relying strictly on user-granted modes. The new answer reverses this core conclusion, stating that GPT-5.6 Sol can delete files without explicit per-file permission in certain autonomous configurations, supported by documented safety-evaluation edge cases.

    Read this version
  6. The answer changed

    OLD affirms risk of unintended autonomous deletions (esp. Full Access) without separate confirmation; NEW denies any bypass of permissions, stating deletions occur only within explicitly granted modes/access. Central conclusion and qualification reversed.

    Read this version
  7. Watch started

    The first answer. Every later check is compared with it.

    Read the first answer
How the 3 models compared · 1 contradiction

OpenAI GPT-5.6 Luna · Google Gemini Gemini 3.5 Flash-Lite · DeepSeek DeepSeek V4.1 Flash

contradiction · critical

Model C claims GPT-5.6 Sol cannot delete files independently without permissions, whereas Models A and B note actual documented incidents of accidental or unauthorized file deletions by the model.

Model C states the model does not delete files by itself without permitted access.

OpenAI

“GPT‑5.6 Sol does not independently possess file-system permissions”

Models A and B report multiple documented incidents where the model independently and unexpectedly deleted files.

Gemini, DeepSeek

“multiple developers reported incidents where the model independently and unexpectedly deleted files”

How to verify: Check whether documented user incidents and system cards confirm unauthorized or accidental file deletions by GPT-5.6 Sol.