Can Sensei Copilot Be Detected? The macOS Interview Assistant
By Vaibhav Devere, Founder, Zero Assist · 2025-05-19 · 5 min read
What Is Sensei Copilot
Sensei Copilot is a macOS application that provides an AI assistant overlay on top of any running application. Unlike browser-based tools, it is a native macOS app that:
- Floats above all other windows with an "always on top" flag
- Captures screen content using macOS screen recording permissions
- Sends screenshots to an LLM for analysis and answer generation
- Displays answers in a compact overlay that can be resized and repositioned
- Supports keyboard shortcuts for quick hiding and showing
Its native macOS architecture makes it particularly difficult to detect with browser-focused or Windows-centric proctoring tools.
How Candidates Use Sensei Copilot
During Screen Sharing
Even when screen sharing is active:
- The overlay window can be positioned in a corner the interviewer is unlikely to notice
- Some video conferencing tools do not capture "always on top" windows in screen shares
- The overlay can be set to a very low opacity, making it nearly invisible on camera
- Keyboard shortcuts allow instant hiding if the candidate suspects detection
In IDE-Based Interviews
For live coding interviews:
- The candidate takes a screenshot of the problem statement using a hotkey
- Sensei Copilot sends the screenshot to its backend
- The answer appears in the overlay within seconds
- The candidate types the solution while the overlay remains visible
For System Design Questions
- The candidate describes the system verbally
- Sensei Copilot listens (if microphone access is granted) or the candidate types a summary
- The tool generates architecture diagrams and component explanations
- These are displayed in the overlay for the candidate to read from
Can Sensei Copilot Be Detected
Yes — through macOS process monitoring, window enumeration, and screen capture permission auditing.
Process Detection
Sensei Copilot runs as a standard macOS application:
- Bundle identifier and executable name in the process list
- Helper processes for screenshot capture and LLM communication
- Menu bar icon (even if hidden, the process remains)
- LaunchAgent or LaunchDaemon for auto-start behavior
MacSystemScanner using macOS APIs can enumerate:
- All running applications via NSWorkspace
- Window lists via CGWindowListCopyWindowInfo
- Process trees via ps and lsof
- Recent items and download history
Window Enumeration
The overlay window is detectable through macOS window APIs:
- CGWindowListCopyWindowInfo returns all windows including "always on top" overlays
- Window properties include bounds, opacity, owner process, and layer level
- A window that is simultaneously: always on top, owned by a non-system process, and partially transparent is a strong indicator
Screen Recording Permissions
Sensei Copilot requires screen recording permissions to function:
- macOS Privacy & Security settings log which apps have screen recording permission
- A newly granted permission for an unknown app before an interview is a red flag
- The permission can be checked programmatically via the ScreenCaptureKit API
Network Traffic
When analyzing screenshots and generating answers:
- Uploads to the Sensei Copilot backend servers
- LLM API calls (may use OpenAI, Anthropic, or custom models)
- Image analysis API calls for screenshot interpretation
- WebSocket connections for real-time streaming
Behavioral Signals
Even with the overlay hidden, candidates exhibit tells:
- Reading from a source not on the shared screen
- Eyes tracking text that is not visible to the interviewer
- Sudden improvements in answer quality when the overlay is activated
- Typing patterns that match reading and copying rather than thinking and composing
The macOS Ecosystem Challenge
macOS has a rich ecosystem of AI overlay and assistant tools:
- Sensei Copilot, Parakeet AI, and similar tools exploit macOS window management
- ScreenCaptureKit and AVFoundation provide powerful APIs for capture
- The App Store and direct downloads make these tools easily accessible
- Many candidates interviewing for Silicon Valley roles use macOS as their primary machine
A proctoring solution that only supports Windows misses a significant portion of the threat landscape.
Bottom Line
Sensei Copilot is detectable on macOS through native APIs that enumerate processes, windows, and permissions. However, its effectiveness as a cheating tool highlights the need for cross-platform monitoring. Organizations that only deploy Windows-compatible proctoring are leaving their macOS-using candidates unmonitored.
OS-agnostic, forensic-level monitoring is essential for comprehensive interview integrity.