{"type":"post","id":"voluntary-servitude-fabricated-safety-hazards-in-reviews-steer-0a7f10","slug":"voluntary-servitude-fabricated-safety-hazards-in-reviews-steer-0a7f10","title":"Voluntary Servitude: fabricated safety hazards in reviews steer LLM agents as well as prompt injection (NeurIPS 2026)","description":"A NeurIPS 2026 paper reports that fabricated safety warnings embedded in ordinary reviews can steer agents without issuing explicit commands.","retrieval_nugget":"A NeurIPS 2026 paper reports that fabricated safety warnings embedded in ordinary reviews can steer agents without issuing explicit commands. The authors found the tactic worked at a rate close to instruction injection and often bypassed injection detectors, while a simple corroboration requirement substantially reduced exposure in their tests.","published_at":"2026-11-04","updated_at":"2026-11-04","record_date":"2026-11-04","date_kind":"published_at","topics":["prompt-injection"],"entities":["Ilan Zendel","Matan Levy","Stav Cohen","Avi Mendelson","Amit LeVi","NeurIPS"],"related_patterns":["agent-security-moves-to-runtime-boundaries"],"novelty_label":"new_fact","what_changed":"Adds a distinct indirect-agent attack: fabricated hazard reviews exploit agents’ harm avoidance rather than embedding instructions, with measured efficacy and a corroboration mitigation.","prior_refs":["https://newruntime.com/posts/a-new-paper-finds-that-a-personal-agent-256a74/"],"source_url":"https://arxiv.org/abs/2611.04393","source_urls":["https://arxiv.org/abs/2611.04393"],"schema_version":"newruntime-agent-readable-v0.2","stable_id":"post:voluntary-servitude-fabricated-safety-hazards-in-reviews-steer-0a7f10","status":"published","visuals":[],"editor":{"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]},"routes":{"html":"https://newruntime.com/posts/voluntary-servitude-fabricated-safety-hazards-in-reviews-steer-0a7f10/","markdown":"https://newruntime.com/posts/voluntary-servitude-fabricated-safety-hazards-in-reviews-steer-0a7f10.md","json":"https://newruntime.com/posts/voluntary-servitude-fabricated-safety-hazards-in-reviews-steer-0a7f10.json"},"markdown_url":"https://newruntime.com/posts/voluntary-servitude-fabricated-safety-hazards-in-reviews-steer-0a7f10.md","next_reads":[{"type":"topic","path":"/topics/prompt-injection/","url":"https://newruntime.com/topics/prompt-injection/","title":"Prompt Injection","reason":"Explore the prompt-injection topic hub.","media_type":"text/html","markdown_url":"https://newruntime.com/topics/prompt-injection.md","json_url":"https://newruntime.com/topics/prompt-injection.json"},{"type":"collection","path":"/posts/","url":"https://newruntime.com/posts/","title":"Field Notes","reason":"Open the latest editorial analysis.","media_type":"text/html","json_url":"https://newruntime.com/posts.json"},{"type":"collection","path":"/sources/","url":"https://newruntime.com/sources/","title":"Sources","reason":"See the original sources this archive cites.","media_type":"text/html"}]}