Detectability & Privacy

How interview tools are (and are not) detected, what interviewers can see, and where no tool is safe.

By Aaron Cao · Updated

The questions in this cluster are the ones candidates Google in the dark, hours before an interview, when they're already nervous: "Will the interviewer know?" "Can Zoom flag this?" "Does Meet record what I see on my screen?"

The honest answer for any real-time AI tool — SubcueAI included — is that detectability depends on architecture, not vibes. A tool that injects into your browser is in a different category from a tool that joins your call as a participant, which is in a different category from a tool that runs as a native desktop overlay outside the meeting client's process space. SubcueAI uses the third architecture for that reason — ScreenCaptureKit on macOS and WASAPI on Windows, with no virtual audio cable and no meeting bot. System audio reaches the AI through OS-level APIs that are not exposed to the conferencing app.

That answers the platform question. The proctoring question is separate. If your interview runs under an explicit anti-AI proctor (LiveHire, certain Mercer Mettl modes, the proctored Bar exam infrastructure), no real-time desktop assistant is safe — those systems are designed to detect exactly what we do, and our responsible use policy says so directly. The answers below try to be specific about which contexts are which.

← All answer topics