Missleading warning message: "Your account was flagged for potentially high-risk cyber activity ..."

Thanks for the bug report. Can you provide more details about how you've configured your model provider? A snippet from your config.toml would be most helpful. That will help us repro the problem. Is this a local model like gpt-oss-120b? Thanks!

Yes, I encountered the warning while using codex with a custom provider where indeed there was a name/slug mismatch.

conig.toml:

model_provider = "openai-responses"
model = "gpt-oss"

[model_providers.openai-responses]
name = "internal"
base_url = "**REDACTED**"
wire_api = "responses"

[features]
remote_models = true

models json descriptor:

    {
      "supports_reasoning_summaries": false,
      "support_verbosity": false,
      "default_verbosity": null,
      "apply_patch_tool_type": "function",
      "input_modalities": [
        "text"
      ],
      "truncation_policy": {
        "mode": "bytes",
        "limit": 10000
      },
      "supports_parallel_tool_calls": false,
      "context_window": 96000,
      "reasoning_summary_format": "none",
      "slug": "gpt-oss",
      "display_name": "gpt-oss",
.
.
.

The problematic message is produced here: https://github.com/openai/codex/blob/5c75aa7b89a70bc2cc410a6fd238749306ec4c5e/codex-rs/core/src/codex.rs#L2565-L2581

/edit: corrected link to problematic source code

OK, thanks. Very helpful!

I am also getting this warning despite being verified.

@iharc-jordan, are you using a custom model provider like the OP in this thread? Or are you trying to use gpt-5.3-codex with the default openai model provider?

@etraut-openai I've tried 0.102.0 version and I use gpt-5.2-xhigh, not codex model and not 5.3. I do have serveral:

⚠ Your account was flagged for potentially high-risk cyber activity and this request was routed to gpt-5.2 as a fallback. To regain access to gpt-5.3-codex, apply for trusted access: https://chatgpt.com/cyber or learn more: https://developers.openai.com/codex/concepts/cyber-safety

and I had passed KYC, but it still flagged that.

Before update to 0.102.0 I never saw that message.

@iharc-jordan, are you using a custom model provider like the OP in this thread? Or are you trying to use gpt-5.3-codex with the default openai model provider?

My bad, this is a default model. I was actually using 5.2 non codex when I got the flag

I got the same issue, I am using 5.2-pro with api key and url https://us.api.openai.com/v1

I got the issue and i am using 5.3-codex-xhigh，please help with the fixing , job needs to be done while i actually have some trouble now , please

If you're not using gpt-5.3-codex and are seeing this message, that's a bug. You can ignore the message. We're working on a fix for this.

I also received this warning after upgrading to codex-cli 0.103.0 and the requested work was "downgraded to codex-5.2"; but I was able to reselect the model on a different task. Using my team account.

You could try deleting ~/.codex and reinstalling—seems like a clean reset sometimes fixes this.

I take it back, installing version 0.101.0 works for me though.

Seems that I'm in the same situation. I got the warning when I was using 5.2 xhigh. But I didn't see it if I was using 5.3codex xhigh.

Using CLI 0.104.0 on Windows. My account was flagged and I cannot complete verification at https://chatgpt.com/cyber. Verification succeeds but immediately restarts 019c7046-7e87-72a0-b448-473c7833fd2c

"Using CLI 0.104.0 on Windows. My account was flagged and I cannot complete verification at https://chatgpt.com/cyber."

Same here. Tried every doc to verify.

My issue (related to custom providers when not returning a model name that exactly matches slug) is fixed by https://github.com/openai/codex/pull/12061

Instead of using the responses API model name, it uses a HTTP header (OpenAI-Model/X-OpenAI-Model). If that header is missing (usually the case with custom providers) the whole check is never made thus the warning isn't emitted anymore.

Closing this; if anyone still has issues please open a new issue and/or track one of the others related to the official codex model providers: https://github.com/openai/codex/issues?q=state%3Aopen%20label%3A%22safety-check%22

Thanks!