Aiselon tracks primary-source data on AI-assisted cheating, assessment fraud, and the tool ecosystems that have made after-the-fact detection architecturally impossible.
The figures below are illustrative estimates we use to frame the assessment-integrity challenge. They reflect the direction and scale that industry observers broadly report, rather than a single formal study. We update our view as credible primary research becomes available.
Estimated year-over-year rise in AI-assisted cheating across online assessments
Illustrative industry estimateOnline exams now face some level of AI cheating risk
Illustrative industry estimateShare of technical-role candidates suspected of undisclosed AI assistance in live interviews
Illustrative industry estimateCandidates say they would use AI in assessments if they believed detection was unlikely — suggesting perceived risk, more than ethics, is the main deterrent
Illustrative industry estimateA defining shift of recent years is not simply that more candidates are willing to cheat — it is that the tools enabling them increasingly operate below the visibility threshold of typical detection systems. Invisible overlays render above an exam without appearing in any screen capture. On-device language models generate answers with no network traffic to inspect. Remote-access proxies route a live exam to a paid expert through encrypted channels indistinguishable from ordinary remote work.
Each of these techniques shares one property: they are invisible to the browser. Application-layer lockdown tools, however well engineered, cannot observe the operating system or the network beneath them. That is the gap Aiselon was built to close. By enforcing integrity at the device and network layer, prevention replaces a detection arms race that the defenders have already lost.