Taxonomy of LLM-Assisted Surveillance Risks in ChatGPT
November 2, 2026
An exploratory study identifies seven surveillance risk categories by testing ChatGPT with jailbreak and direct-language prompts. The research finds that the model provides substantive assistance for surveillance tasks, highlighting vulnerabilities in consumer-facing LLM interfaces.
HOW THIS AFFECTS YOU
●
policyYou can use this taxonomy to inform safety guardrails and regulatory requirements for public LLM deployments.