Smoke Codex failing: OpenAI cybersecurity restriction now affecting codex smoke test

[github-actions]

Summary

Smoke Codex is now failing with the same OpenAI cybersecurity restriction that has been blocking AI Moderator since 2026-02-26. This represents an expansion of scope — the restriction is now affecting a second workflow.

Error message:

This user's access to this model has been temporarily limited for potentially suspicious activity related to cybersecurity. Learn more about our safety mitigations: (platform.openai.com/redacted)

Evidence

• Run §22676642094 — Smoke Codex [smoke-detector] 🔍 Smoke Test Investigation - GenAIScript OPENAI_API_KEY Missing #2142 FAILED (2026-03-04)
• Agent job failed; all 6 api.openai.com requests returned the cybersecurity restriction error
• AI Moderator has been failing with the same error since 2026-02-26 (issue [aw] AI Moderator failed #18922)

Impact

• Smoke Codex is the primary smoke test for the codex engine (OpenAI Codex / GPT-5.x-Codex)
• With both codex-engine workflows blocked, there is no visibility into codex engine health
• Reactive moderation (AI Moderator) also remains down

Recommended Actions

1. Immediate: Investigate whether the OpenAI API account restriction can be lifted (contact OpenAI support)
2. Short-term: Consider switching Smoke Codex to use a different model (e.g., gpt-4.1 or Claude) as fallback
3. Short-term: Update AI Moderator to use an alternative non-codex model (see [aw] AI Moderator failed #18922)
4. Track: Monitor whether restriction expands to other workflows

Related

• Issue [aw] AI Moderator failed #18922 — AI Moderator failing (same restriction, day 7+, expires 2026-03-07)
• This run: §22681140788

Generated by Agent Performance Analyzer - Meta-Orchestrator · ◷

• expires on Mar 6, 2026, 5:40 PM UTC