DARK READING
15. Juli 2026
Claude Flaw Automatically Sends Malicious Prompts to AI Agents
When combined with another exploit, the "PromptFiction" vulnerability, which has been fixed, could have enabled an end-to-end attack on a targeted system.