Hackers redundant? AI increasingly attacks independently: ‘It’s a rat race’
Google researchers have discovered a fully autonomous AI system designed to conduct cyberattacks without human supervision. The software exploits unpatched software vulnerabilities at high speeds, raising concerns about a new arms race in cybersecurity as malicious actors adapt open-source models for illicit purposes.
Google's cybersecurity researchers have identified a fully autonomous AI system capable of conducting cyberattacks without human intervention. This software scans for unpatched security vulnerabilities in organizational systems and launches attacks within minutes of discovering an opening. Evidence, including Russian-language notes, was found on a server targeting media, entertainment, and education websites in France, though the origin of the attack remains unknown. Experts warn that the race to secure networks is intensifying as AI accelerates the speed and efficiency of such digital breaches.
Ethical hacker Joost Grunwald notes that while AI-driven security tools are common, malicious actors are increasingly removing ethical safeguards from open-source software, particularly those based on Chinese models. By stripping away these guardrails, hackers can easily generate phishing content and orchestrate automated strikes. For large organizations, this creates a constant race against time, as they must secure thousands of potential entry points before automated AI systems can exploit them. The rise of censorship-free, malicious AI models has significantly lowered the barrier to entry for cyberattacks, forcing companies to reconsider their digital defense strategies.
Based on reporting by ad. Translated and condensed by LocalHeadlines.



