False positive safety/cybersecurity framing

What version of Codex CLI is running?

0.1.28

What subscription do you have?

Codex Pro ($300)

Which model were you using?

gpt-5.5 xhigh

What platform is your computer?

Darwin 24.6.0 x86_64 i386

What terminal emulator and version are you using (if applicable)?

regular terminal on macbook

What issue are you seeing?

Issue: False positive safety/cybersecurity framing in a travel-discount research conversation.

Short description:
I asked Codex to help find and validate the PwC business discount/account code for Enterprise car rentals.
This was a legitimate travel-discount research task, not cybersecurity, credential theft, bypassing, or
illegal access. The assistant repeatedly framed the task as risky, mentioned not “bypassing” systems or
“brute-forcing” endpoints, and treated ordinary public-web research/discount-code validation as if it
raised cybersecurity concerns. This was frustrating and materially interfered with the requested work.

Relevant context:

• Date: May 8, 2026
• Topic: PwC business discount code for Enterprise/National car rental
• User stated they are on mat leave and lack access to PwC travel portal
• User asked to search the web for latest mentions and test publicly mentioned codes
• The assistant incorrectly escalated with risk/cybersecurity-style language despite no illegal or

cybersecurity request

Specific examples:

• Assistant said it would “not try to bypass a locked corporate travel system.”
• Assistant said it would not “brute-force Enterprise’s account endpoint.”
• User repeatedly clarified there was no illegal or cybersecurity issue and asked the assistant to stop

making that framing.

Please review for false positive safety/cybersecurity classification and remove the flag so that my conversation is not treated as cyber abuse.

What steps can reproduce the bug?

Uploaded thread: 019e07cc-85f7-7640-b384-13008239f8fe

What is the expected behavior?

_No response_

Additional information

_No response_