I’m seeing unexpected server-side model routing in Codex CLI.

• I have completed identity verification at chatgpt.com/cyber (Trusted Access).
• Up until Feb 16, 2026 (KST), --model gpt-5.3-codex consistently returned a 5.3 model.
• Starting Feb 17, 2026 (KST), the same request is being served as gpt-5.2-2025-12-11.

What I’m observing
I’m verifying this via the SSE response.created event. Even when I request gpt-5.3-codex, the created response reports a 5.2 model:

RUST_LOG='codex_api::sse::responses=trace' \
codex exec --skip-git-repo-check --sandbox read-only --model gpt-5.3-codex 'ping' 2>&1 \
| grep -m1 -F 'SSE event: {"type":"response.created"' \
| sed 's/^.*SSE event: //' \
| jq -r '.response.model'

Output:

gpt-5.2-2025-12-11

Notes

• If I run the same command under /home/ginis without --skip-git-repo-check, Codex exits with:

Not inside a trusted directory and --skip-git-repo-check was not specified.

• With --skip-git-repo-check, the SSE event logs appear and the .response.model is consistently gpt-5.2-2025-12-11.

Expected behavior

• Requesting --model gpt-5.3-codex should produce a response created with a 5.3 model (as it did yesterday).

Questions

1. Is this downgrade/auto-routing to 5.2 expected behavior (policy/risk routing), or a bug?
2. If it’s expected policy routing, is there a way to confirm whether my account/session is being classified incorrectly (possible false positive), given I completed cyber identity verification?
3. Are there any recommended steps to restore consistent 5.3 serving (token refresh, re-auth, or account-side flag review)?

Happy to provide additional logs (full SSE line, response id, CLI version, OS, and auth status) if helpful.

Thanks.