AI Problems Index
Back to Real AI Issues
Critical

AI-enabled hacking

GPT-4 autonomously exploited 87% of one-day CVEs; lowers attacker skill bar.

Description

Advanced AI models like GPT-4 have demonstrated the ability to autonomously exploit 87% of one-day Common Vulnerabilities and Exposures (CVEs), significantly lowering the skill barrier for potential attackers.

Rebuttals to Common Fallacies

  • While AI can be used defensively too, the asymmetry of attack vs. defense means this remains a serious concern.
  • The democratization of hacking capabilities through AI requires new approaches to cybersecurity beyond traditional methods.