Bad Memories Still Haunt AI Agents

Cisco found and fixed a significant vulnerability in the way Anthropic handles memories, but experts warn that mishandled memory files will continue threaten AI systems.