Experts Warn of AI Agents as Potential Threats

Experts from Google, Meta, and universities warn AI agents may be unreliable.

Experts from Google, Meta, and several universities have urged that AI agents be considered unreliable systems. Researchers have suggested implementing protective mechanisms across the entire IT infrastructure, not just within language models. They believe that separating data and instructions, along with restricting access rights, will help prevent cryptocurrency wallet hacks and information leaks.