Research News › Archive › Technology & AI
How attackers persuade AI agents to break the rules
Technology & AI · 19 August 2026
19.08.26 - As AI assistants evolve into AI agents, a new EPFL study has found that the biggest safety risks in their use may not come from single malicious prompts, but from carefully orchestrated conversations. Today, most of us interact with AI assistants -
Find the funding to take work like this further
Global Research Partnerships puts the world's major public funders in one search, and connects you with the collaborators who help you win. Free to search, always.
Search research funding →More research news
University of Michigan
U-M researchers use AI to help set record-setting sensitivity in search for new physics
MIT
When AI art has no author: Study finds generated images often can’t be traced to training data
University of Michigan
Premium customer, quickest service? Not so fast, at least in world of agentic AI-driven services
University of Washington
Q&A: UW professor Hossein Naghavi uses terahertz waves to help sensors augment human vision
National University of Singapore
NUS makes AI courses compulsory, gives free access to ChatGPT Edu
Carnegie Mellon University