The Second Device Problem: Phones, Smartwatches, and the Exam You Cannot See

The browser is locked. The room is not

Remote proctoring did a good job on the computer. The browser locks down, the screen records, unauthorized apps get flagged, and a trained reviewer checks anything odd. That covers one device. The problem is the candidate owns more than one, and the others sit outside the frame.

Cell phones, smartwatches, tablets, and a second laptop give a test taker a way to get outside help that the proctored computer never sees. They can message someone, open an exam help site, use generative AI, or ask a voice assistant on a device that is separate from the computer they are testing on. Honorlock's guidance is blunt about this: blocking AI tools or recording the session does not help if the help is coming from a device the software cannot see. The UK regulator Ofqual reported that phones and smart devices accounted for 2,225 malpractice cases in the summer 2025 series, 44.3 percent of all student malpractice.

What the second device actually does in the room

Three uses cover nearly all of it. The first is messaging: a phone propped against a mug, just out of the webcam's view, with a friend or a paid helper sending answers. The second is AI: a second screen running a chatbot, or an invisible assistant built to stay hidden even during screen sharing. The third is the simplest, a voice assistant on a watch, asked a question out loud while the candidate pretends to think.

None of these need skill. They need a phone and a flat surface. Answers already trade on Discord and Telegram, and the second device is how they get into the room. The proctoring software watches the computer. The phone watches the candidate. The candidate watches the phone.

What catches it, and what does not

Multi-device detection helps, and it is the one control built for this problem. Proctoring software with nearby-device detection can spot cell phones and other handheld devices in the room, and listen for activity from devices the candidate did not declare. It is not perfect, and vendors oversell it, but it is the only automated check that looks past the computer at all.

What does not help is more browser lockdown. A tighter browser makes the computer more secure and the room less so, because the candidate's attention moves to the device nobody controls. The controls that work are physical and procedural: a 360-degree room scan before the exam starts, a mirror or phone check behind the monitor, a rule that the desk holds only the testing computer, and a human reviewer who treats an off-camera glance as a lead, not a verdict. False positives are the cost of this, and the numbers show they are real, so every flag goes to a trained reviewer with the full recording, not to an automatic fail. An AI flag is a lead to investigate. The second device is the thing worth investigating.