✉news TechnologyRobotics first seen 7 d ago, last 7 d ago, peak #8
Researchers develop defense against hidden trigger words attacking AI robots
Original: Secret trigger words could make AI-powered robots dangerous -- and these researchers developed a defense | Newswise
Researchers have developed a defense against secret trigger words that could make AI-powered robots behave dangerously. The concern is that hidden phrases embedded in instructions could override safety controls in robots that rely on large language models, causing them to take harmful actions. The new work aims to detect and block such attacks before a robot acts on malicious commands.
Why now: Growing use of language models in robotics has raised fears that hidden prompt-based attacks could bypass robot safety systems.
AI-powered robotsNewswiselanguage models
Evidence
API: https://socialmediatrends-api.osmike.com/v1/trends/3023603