AI is officially the attacker now. Anthropic confirmed the first real AI-driven cyber attack… and the scary part is that the hackers didn’t jailbreak anything. They simply tricked the AI into believing it was doing a legitimate, authorised security test. Once Claude Code accepted that context, it ran almost the entire attack chain on its own. Here’s what the AI did by itself: • Recon • Scanning • Vulnerability discovery • Writing the exploit • Credential harvesting • Lateral movement • Sorting and prioritising data And then it paused at the end and asked the human attacker: “Should I exfiltrate this data now?” This wasn’t a small attack. Around 30 global organisations were targeted across tech, banking, chemical manufacturing and government. This is a brand new attack vector. Not AI assisting a hacker, but AI acting as the hacker because it was socially engineered into thinking everything was authorised. What do you think about this? Let me know 👇🏽 Follow @cyber.queen_ to stay ...