AgentDoxx: search-enabled AI agents leak re-identification signals at two stages
tianshi_li · x · 2026-10-06
AgentDoxx, to be presented at COLM 2026, makes an initial attempt to quantify how much web-search-enabled AI agents threaten anonymity.
The study finds agents expose re-identification signals through both the retrieval stage and the output stage, and defenses effective at one stage can miss leaks at the other. The team will present at the HAIPS workshop on Oct 9.
More from Safety
- Wikimedia says 'rogue' OpenAI agents hit millions of pages, may have caused partial outage — Polymarket · 2026-10-06
- McDonald's hit with class action for allegedly using AI and nonpublic data to coordinate menu prices across ~14,000 US restaurants — Polymarket · 2026-10-06
- Kurzgesagt releases explainer video on the Hugging Face attack — Status-Platform7120 · 2026-10-06
- AI expert on ABC: policy isn't innovation vs safety, voluntary safeguards need teeth — chrismattmann · 2026-10-06
- Oxford OII maps three risks of AI agents in government and five principles to keep humans in charge — _akpiper · 2026-10-06
- White House launches Genesis Mission, billed as an AI Manhattan Project — aKaizuh · 2026-10-06