Alvy AI Proctoring Agent — agentic threat model

Uses advanced LLMs to analyze behavior. Highly vulnerable to adversarial prompt injection or evasion techniques where students manipulate their environment or inputs to bypass detection or force false negatives.

Not certain from the listing — likely processes real-time video, audio, and screen capture data. Threats include unauthorized data exfiltration of sensitive student biometrics and lack of clear data retention/poisoning protections.

Not certain from the listing — orchestrates browser-blocking and device-detection tools. Threats include insecure tool integration where the agent's system-level hooks could be exploited by local users to bypass security controls.

Not certain from the listing — likely deployed via web browsers or local client installations. Threats include local tampering with the client-side agent to feed spoofed camera/audio streams or disable the monitoring process.

Not certain from the listing — claims 95% accuracy but lacks transparent logging or human-in-the-loop verification details. Threats include evaluation gaming by students and lack of explainability for disputed cheating flags.

Not certain from the listing — handles highly sensitive student data and academic records. Lack of explicit compliance alignment (e.g., FERPA, GDPR, COPPA) poses significant regulatory and privacy risks.

Not certain from the listing — operates primarily as a standalone proctoring tool. Threats are limited to integration points with Learning Management Systems (LMS) where compromised LMS credentials could abuse the proctoring agent's trust.