ScrapeBench: evaluating the legal compliance of agent website scraping
Can AI Agents Be Trusted to Follow the Rules?
AI summary
Multiple studies evaluate the legal compliance of AI agents: ScrapeBench proposes the first benchmark testing their adherence to website scraping policies, and Authorization Drift links agent out-of-bounds behavior to data minimization provisions in four jurisdictions. A systematic review examines 3,184 documents, identifies 95 published scenarios and groups them into eight categories; another study examines AI law-abiding attitudes and the compliance evidence tension in the technical translation of the EU AI Act.
Related papers