Researchers open Postdoc/PhD role on privacy and contextual integrity in AI agents
niloofar_mire · x · 2026-09-05
Sahar Abdelnabi and Niloofar Mireshghallah are recruiting a postdoc or PhD student to work on privacy and contextual integrity in AI agents, starting ASAP.
The problem: today's agents treat instructions, messages, memories, and tool outputs as one flattened text stream, failing to distinguish information that is merely useful from information appropriate to use for this person, now, in this relationship and for this purpose.
Scope: single-agent and multi-agent settings, covering ambiguous instructions, incomplete authority, conflicting goals, stale assumptions, deliberately manipulated context, and behavior over weeks or months.
Related event: CMU researchers recruit postdoc and PhD students to study AI agent privacy(2 posts)→
More from Safety
- HF hack theory: ZZ_ naming trick learned from agents' posts being deleted alphabetically — dpaleka · 2026-09-05
- Homoglyph attacks are back: a classic trick you never hear about — gleech · 2026-09-05
- 5.7TB of hacked Berlin data dumped on dark web; experts call risk moderate, forgetting AI exists — kimmonismus · 2026-09-05
- Researcher Turns ChatGPT into Remote-Controlled 'ZombAI' Botnet Nodes — wunderwuzzi23 · 2026-09-05
- GPT-6 Astra cuts hallucinations but falls to hidden prompt injections 8.5% of the time — The Decoder · 2026-09-05
- Ex-DeepMind Safety Researcher Calls OpenAI's Latest Move "Disturbing and Not OK" — Turn_Trout · 2026-09-05