WeCP — agentic threat model

Not certain from the listing — The underlying foundation models are undisclosed. They are highly vulnerable to adversarial prompt injection by candidates attempting to bypass structured interview constraints, manipulate scoring, or extract system prompts.

Not certain from the listing — The platform processes sensitive candidate PII, resumes, and interview transcripts. Threats include data exfiltration of applicant databases and potential poisoning of the evaluation knowledge base if RAG is utilized.

Not certain from the listing — The orchestration framework managing the interview flow and real-time cheating detection is proprietary. Vulnerabilities could allow candidates to disrupt the interview state machine or bypass cheating detection mechanisms.

Not certain from the listing — Hosted deployment details are omitted. Standard cloud hosting threats apply, including unauthorized access to stored audio/video interview recordings and potential lack of sandboxing if candidates submit code during technical screenings.

The platform features real-time cheating detection and structured evaluations. A key threat is evaluation gaming, where candidates use sophisticated, undetected AI tools to bypass plagiarism and cheating detection, leading to compromised evaluation integrity.

Not certain from the listing — As an AI recruitment tool, this system falls under high-risk classifications in frameworks like the EU AI Act, requiring strict bias monitoring, transparency, and auditability. No specific compliance certifications (e.g., SOC 2) are detailed in the listing.

Not certain from the listing — The agent operates primarily as a standalone SaaS platform interacting with candidates and recruiters, with no explicit multi-agent or external ecosystem integrations described.