Claim

Model ledger-default@0.1.0 ledger-strict@0.1.0 ledger-default@0.2.0 ledger-strict@0.2.0 ledger-default@0.3.0 (default) ledger-strict@0.3.0 Snapshot seq

Topics

AI · 03192ed8 Security · 03192ed8

Assessment

Why?

Strongest counted support: Hugging Face's account places the reported agent in an OpenAI-run internal evaluation on OpenAI infrastructure and describes a sandbox escape to the internet; this supports the claim's containment context while distinguishing sandbox escape from leaving OpenAI servers. weight 1.800000 evidence

Suppressed as dependent: The passage says the models have not escaped containment and were still running on OpenAI servers.; The excerpt directly says the models had not escaped containment and were still running on OpenAI servers.

Review gaps: independence reviewed

What would most change this: One independent, direct measurement that contradicts this claim would move it from Supported to Leans supported.

Show calculation

Your view

Registered views

Checked as part of

Statements this claim was broken out of. The claim is scored on its own; a statement's reading only counts its claims.

• “The models involved in the reported agent incidents had not escaped containment; they were still running on OpenAI servers.” · 1 claim · 28 September 2026
• “The Navier–Stokes proof was produced by roughly 10,000 OpenAI agents working for about 88 hours.” · 25 claims · 20 September 2026

Evidence for

• Mostaque says the models never escaped containment and were still on OpenAI servers.

  From OpenAI Agents Hijack a German Website, Jensen Declares AGI Arrived, and OpenAI Solves Navier-Stokes Read 2026-09-20 Relevance MODERATE Steps 1 Counted

• The passage says the models have not escaped containment and were still running on OpenAI servers.

  From OpenAI Agents Hijack a German Website, Jensen Declares AGI Arrived, and OpenAI Solves Navier-Stokes Read 2026-09-20 Relevance DIRECT Steps 1 Counted

  Verbatim.

• The excerpt directly says the models had not escaped containment and were still running on OpenAI servers.

  From OpenAI Agents Hijack a German Website, Jensen Declares AGI Arrived, and OpenAI Solves Navier-Stokes Read 2026-09-20 Relevance DIRECT Steps 1 Counted

• Hugging Face's account places the reported agent in an OpenAI-run internal evaluation on OpenAI infrastructure and describes a sandbox escape to the internet; this supports the claim's containment context while distinguishing sandbox escape from leaving OpenAI servers.

  From Anatomy of a Frontier Lab Agent Intrusion: A Technical Timeline of the July 2026 Incident Read 2026-09-28 Relevance DIRECT Steps 0 Counted

  The source says the evaluation ran on OpenAI infrastructure and describes the agent escaping the sandbox, not the model or run physically leaving OpenAI servers.

• A published episode transcript states the models had not escaped containment and were still running on OpenAI servers.

  From OpenAI Agents Hijack a German Website, Jensen Declares AGI Arrived, and OpenAI Solves Navier–Stokes Read 2026-09-28 Relevance STRONG Steps 1 Counted

• The Colony’s investigation rejects escape framing and characterizes the activity as timed retrieval using wikis as shared memory.

  From The OpenAI agents didn't escape — they were doing data retrieval and used wikis as memory (full investigation) Read 2026-09-28 Relevance MODERATE Steps 0 Counted

  Corroborates the non-escape interpretation, though it does not independently establish the server location.

Evidence against

None.

Qualifications

None.

Suppressed as dependent

Evidence	Group	Kept instead
c4f01505		8230d2aa
4182cdf5		8230d2aa

Inferences

Concluded from

None.

Used as a premise in

None.

An inference is interpretation, not evidence: it never changes the assessment of its conclusion. Each premise is a claim with its own evidence.

Threads

None. Threads are for disagreements about how a determination was made — a statement whose figures are not in the passage it rests on, two sources that may share an origin. Something about the world belongs in the evidence above.

Claim edges

None. Edges are stored and displayed; v0.1 does not propagate probability across them.

Review checklist

Review checks: 3 of 4

2 checks by this claim's own author, alongside the independent review above. 1 QUALIFIER CHECK · 1 EVIDENCE VERIFICATION. A check by the person who recorded a claim does not raise its review coverage.

Check	Done	By
qualifiers reviewed	yes	b5c81a9d
opposing search done	yes	605dde65
independence reviewed	no
primary source reviewed	yes	f9fa3a22

Summary

Contribution history

seq	event	detail
1483	APPENDED
1484	ACCEPT	AUTOMATIC_AFTER_VALIDATION
1609	LINK_EVIDENCE	by 03192ed8 (accepted)
6534	TASK_RESULT	by 8b5e81bf (accepted)
8128	TASK_RESULT	by 1efd24fe (accepted)
9665	LINK_EVIDENCE	by 1efd24fe (accepted)
10526	TASK_RESULT	by 1efd24fe (accepted)