Google Warns of Industrial Scale AI-Based Hacking

Alphabet's Google (GOOGL.O) reported on Monday that hackers from a well-known cybercrime group utilised artificial intelligence to find a software vulnerability and an exploit to exploit it for the first time.

According to a report from Google's Threat Intelligence Group, a popular open-source system administration tool was the target of the intended attack, but it was stopped before it could be used as part of a “mass exploitation event.”

This is the first time Google has discovered attackers utilising AI to find a new vulnerability and try to exploit it on a large scale.

AI-Driven Attacks Increase

According to John Hultquist, chief analyst at Google Threat Intelligence Group, the results are probably just the “tip of the iceberg” in terms of how state-sponsored hackers and criminals are advancing AI hacking.

Additionally, the report described how hackers are starting to delegate some of their cyber operations to artificial intelligence, which they use to create malware and search for software bugs on their own.

As attackers start using AI systems not only as research tools but also as active components that can analyse targets, produce code, and make decisions with little human oversight, the shift represents an early step toward more autonomous cyber operations, according to the researchers.

Global Alert on AI Threats: Geopolitics Fuels Cyber Risk

The discoveries come as governments worldwide struggle to control potent new AI models that may make it simpler for hackers to identify targets and conduct attacks using both known and recently found software faults.

Moreover, the results are consistent with recent cautions issued by European financial regulators, who claim that at a time of increased geopolitical tensions, quickly developing AI models are boosting the speed and scope of cyber dangers.

According to the report, state-affiliated hacking groups connected to North Korea, China, and Russia, as well as cybercriminals, are already experimenting with directly incorporating AI into attack workflows.

Google cautioned that although the approaches are still in their early stages, they could speed up cyber assaults by lowering the time and knowledge required to launch sophisticated attacks.

AI Hacks Set New Records

Attackers tried to take advantage of the vulnerability in a method that would have circumvented two-factor authentication, which is an essential last line of defence for many businesses.

Remarkably, Google simply confirmed that it wasn't Gemini, refusing to reveal the hacker organisation, the targeted software, or the AI model that was employed.

Furthermore, the report notes that groups connected to China and North Korea are becoming more interested in automated vulnerability detection technologies like OpenClaw.

Beyond this incident, researchers discovered VillainNet, a hidden backdoor in AI for self-driving cars, and showed that AI models can be remotely reverse-engineered via walls using a tiny antenna.

In another instance, Discord users were able to view Anthropic's limited Mythos model by circumventing access barriers. Although it is still in its infancy, a new area of AI pentesting is developing on the defensive side to stress test model behaviour.

Read more news:

• Australia Approves New Gas Field Near the Twelve Apostles to Enhance Supply
• Oil & Petrol Prices Decrease Substantially, Fuelled by Optimism that the Strait of Hormuz Will Reopen
• Earth Day 2026: Who Can Actually Save the Earth Now?