AI Cyber Attacks: New Threat Looms, Security Firms Warn

A new generation of AI-driven cyber attacks is on the horizon. Security firms are testing advanced AI models, finding them "incredibly good" at discovering vulnerabilities and generating exploits. Palo Alto Networks, among others, is testing Anthropic's Mythos AI and OpenAI's latest models. While access is currently restricted, similar capabilities are expected to be widely available within three to five months. This widespread availability will significantly disrupt IT security. These AI models can help address security flaws, but they can also become powerful cyber weapons. Teaching AI to write secure code also teaches it to find insecure code and vulnerabilities. This means a deluge of patches for existing software, leading to more unpatched systems and increased opportunities for attackers. There will also be an increased risk of malware attacks delivered through software updates, especially in open-source environments. This development could place significant new stress on organizations already struggling with security threats.