Once Feared, AI Now Used by Hackers for Cyberattacks

The emergence of advanced Artificial Intelligence (AI) programs capable of autonomous simulation, such as Claude and Mythos, has raised cybersecurity concerns. Experts warn that AI could be misused by hackers for cyberattacks. However, new findings indicate these concerns are now materialising. Hackers are reportedly using AI to identify and exploit vulnerabilities in specific software. ‘For the first time, GTIG has identified cybercriminals using zero-day exploits we believe were developed with AI,’ Google stated in its official report. GTIG explained the exploit targeted a popular open-source web-based system administration tool, whose name was not disclosed. Through the AI-bypassed vulnerability, hackers can circumvent two-factor authentication (2FA) to access victim systems. This allows hackers to take control of user devices and steal data. Despite the AI-enabled breach, Google confirmed the attack was thwarted before mass exploitation occurred. Google claims it alerted the relevant software developers to implement immediate mitigation and patch the security flaw. Google is confident the attack, comprising a Python-based exploit chain, was created with the assistance of a large language model (LLM), often considered the ‘brain’ of AI products. Researchers also discovered fake CVSS (vulnerability) scores generated by AI ‘hallucinations’ within the code. Alongside the coding patterns, the nature of the vulnerability further supports the suspicion of AI involvement. This is because the exploited flaw is a high-level semantic logic bug, which AI is better equipped to detect than traditional methods such as fuzzing or static code analysis. So, which AI was used by the cybercriminals? Google has not disclosed the specific model.