AI-assisted exam security combines automated detection of potential risks with trained human review so programs can spot emerging threats at scale without turning algorithmic flags into final decisions. It strengthens test integrity, reduces cheating opportunities, and protects credential value while avoiding the pitfalls of AI-only proctoring.
Across education and professional testing, exam security is under new pressure. Large-scale research from Wiley, summarized by Integrity Advocate, surveyed more than 2,900 students and instructors across North America and found that cheating is measurably increasing online, with 86% of instructors saying students are more likely to cheat in online courses than in person and nearly half of students reporting they use generative AI in their coursework (Integrity Advocate / Wiley research). Even for in-center exams, program owners know the same technologies and behaviors can spill into the testing room.
At the same time, AI-only proctoring has shown real limits. Integrity Advocate describes common cases where automated systems flag ordinary behavior—like looking away from the screen to think, moving lips while reading, or disability-related behaviors—and treat those flags as misconduct when no human ever reviews the session (Integrity Advocate). The result is false positives and decisions programs struggle to defend to candidates, regulators or employers.
Research presented at the 2026 CHI Conference on Human Factors in Computing Systems shows that candidates’ acceptance of proctoring has increased since 2020, but largely because they have become resigned to privacy tradeoffs and appreciate the convenience of exams, not because their concerns disappeared (CHI 2026 proctoring study). For high-stakes programs, that is a warning sign: pushing surveillance too far or leaning entirely on opaque algorithms can erode trust in the long term.
ProGuardian responds to this environment by keeping humans in the loop. AI helps detect potential issues quickly and consistently at the workstation level, but a trained security specialist always reviews each flag before anything reaches on-site staff. That design acknowledges what the data shows: AI is powerful at spotting patterns, but only people can interpret context, consider accommodations and make defensible decisions.
ProGuardian adds a new, scalable layer of workstation-level security across Prometric’s global test center network. Its planned deployment covers thousands of environments, with more than 26,000 additional cameras across 13,000+ workstations in over 500 cities and 100+ countries. That reach matters for programs that test candidates worldwide and need consistent standards wherever exams are delivered.
At the individual workstation, ProGuardian starts by capturing a more complete view of the testing environment. Two cameras at each desk monitor the test taker, their workspace and the surrounding area for the duration of the exam, augmenting existing overhead room surveillance. This dual-camera setup helps reveal risks that a single, room-level camera might miss, such as a small device positioned just outside the primary field of view.
During the exam, AI-assisted detection analyzes the video feed in real time for potential prohibited items like phones, smartwatches or covert recording devices. If the system spots something that matches configured risk patterns, it raises a flag. Crucially, that flag is not treated as proof of misconduct. Instead, it becomes a prompt for expert human review.
Prometric’s 24/7 Security Command Center receives every flag and reviews it against your program’s policies and approved accommodations. A specialist can see exactly what was captured, the timing, the candidate’s behavior, and any relevant notes about permitted items. For example, if a candidate has an approved medical device or assistive technology, that context is considered explicitly so accommodations are never misinterpreted as violations.
Only concerns that trained experts validate are escalated to on-site Test Center Administrators for further analysis and potential action. This means center staff are not overwhelmed with noise from AI, and candidates are not interrupted based on unreviewed alerts. Incidents are documented, time-stamped and tied to the relevant evidence, supporting thorough post-exam investigation when needed.
Stronger security is only useful if it also protects program reputation, candidate trust and legal compliance. ProGuardian is designed as part of Prometric’s end-to-end security approach—covering identity verification, secure exam delivery, monitoring, investigation and response—with privacy, fairness and responsible oversight built in from the start.
Privacy-by-design means security controls are planned with data protection in mind, instead of bolted on after deployment. ProGuardian was developed with Prometric’s Data Protection Officer at the table, so questions like data minimization, retention, access control and regional regulatory requirements are addressed from the outset. This is especially important when video from thousands of workstations is processed across more than 100 countries, each with its own expectations and rules.
Fairness is supported by the commitment to “no action without human review.” Every AI flag is examined by a trained specialist before it can affect a candidate. That oversight reduces the false positive problem documented in AI-only proctoring scenarios and helps ensure that ordinary behaviors, cultural differences, or disability-related movements are not misclassified as cheating. Programs gain stronger evidence of integrity without sacrificing accessibility.
Shared standards across the Prometric network further enhance defensibility. Test centers follow the same core security protocols, adapted only where local regulations or client policies require it. When an incident occurs, programs can point to a consistent, documented process: workstation-level capture, AI-assisted detection, expert review and on-site response. That structure supports defensible decisions if results are challenged by candidates, employers or regulators.
Ultimately, what is at stake is more than protecting individual test items. It is the value of the credential itself and the decisions it informs—who gets licensed, hired or admitted. By investing ahead of emerging threats with tools like ProGuardian, programs can stand behind each result with confidence, knowing that every exam was protected by both advanced technology and accountable human judgment.