Google's Threat Intelligence Group (GTIG) reported that it successfully prevented a hacking attempt that aimed to exploit artificial intelligence models for a large-scale cyberattack. The hackers were attempting to utilize an AI model to identify and exploit a zero-day vulnerability, which is a software flaw that is not yet known to developers, allowing them to bypass two-factor authentication.
Google stated it has 'high confidence' in its findings but did not reveal the identity of the hacker group involved. Importantly, Google clarified that its own AI model, Gemini, was not implicated in this incident.
This situation highlights the increasing sophistication of cybercriminals who are leveraging AI tools, such as OpenClaw, to exploit software vulnerabilities, posing significant risks to companies and government entities despite substantial investments in cybersecurity.
The report also noted that hacker groups from China and North Korea are particularly interested in using AI for vulnerability discovery. This incident follows previous concerns in the industry, such as Anthropic's decision to delay the rollout of its Mythos model due to fears that it could be misused for identifying software vulnerabilities.
Anthropic has since made the model available to select testers, including major firms like Apple and Microsoft. Additionally, OpenAI recently announced a limited preview of GPT-5.5-Cyber, a version of its AI model designed for cybersecurity teams. Overall, this event underscores the evolving landscape of cyber threats and the critical need for robust defenses against AI-enabled attacks