Adversaries hijacked AI security tools at 90+ organizations. The next wave has write access to the firewall

Adversaries injected malicious prompts into legitimate AI tools at more than 90 organizations in 2025, stealing credentials and cryptocurrency.