Google says cybercriminals and state actors are rapidly weaponising Artificial Intelligence
Artificial intelligence has already transformed cybercrime into an industrial scale threat, according to a chilling new warning from Google researchers who say hackers are now using advanced AI systems to accelerate and refine attacks at unprecedented speed.
The report from Google’s threat intelligence group claims criminal organisations and state linked actors from countries including China, North Korea and Russia are increasingly exploiting commercial AI tools to improve cyber operations.
Researchers said hackers are using systems such as Gemini, Claude and tools developed by OpenAI to strengthen malware, identify software weaknesses and dramatically increase the scale and sophistication of attacks.
“There’s a misconception that the AI vulnerability race is imminent,” warned John Hultquist, chief analyst at Google’s threat intelligence group. “The reality is that it’s already begun.”
According to the report, cybercriminals are no longer simply experimenting with artificial intelligence. Instead, AI has rapidly become embedded inside active hacking operations. Researchers say threat actors are using the technology to test attacks, maintain long term access to systems and automate tasks that previously required large teams of skilled hackers.
The warning comes amid growing global fears over how powerful modern AI systems have become, particularly in coding and vulnerability detection. Some experts now believe AI could fundamentally reshape the balance between attackers and defenders in cyberspace.
One of the most alarming developments involved the AI company Anthropic, which reportedly refused to release one of its newest models, Mythos, after concluding the system posed a major security threat if misused.
Anthropic claimed Mythos had discovered previously unknown “zero day” vulnerabilities across major operating systems and web browsers. Such flaws are especially dangerous because developers remain unaware they exist, leaving systems exposed to attack.
The company warned the discoveries required “substantial coordinated defensive action across the industry” before the technology could safely be released.
However, Google’s findings suggest dangerous AI assisted cyberattacks are already happening even without Mythos becoming publicly available.
Researchers revealed that one criminal group recently appeared close to launching a “mass exploitation” campaign using a zero day vulnerability discovered with the help of another AI large language model.
The report also highlighted the growing popularity of OpenClaw Github, an AI tool that attracted widespread attention earlier this year for allowing users to hand over major parts of their digital lives to autonomous AI agents without significant safety restrictions.
Security experts believe these developments could dramatically change the future of hacking.
Steven Murdoch said AI systems are now becoming deeply integrated into cybersecurity work on both sides of the conflict. He argued that while hackers are clearly benefiting, defenders can also use AI to strengthen protection systems and identify weaknesses faster.
Still, concerns are intensifying over whether governments, businesses and public institutions are prepared for the speed at which AI driven cyber threats are evolving.
The debate extends beyond cybersecurity alone. A separate report from the Ada Lovelace Institute questioned optimistic claims about the economic benefits of AI adoption, warning that many projections surrounding productivity gains rely on uncertain assumptions and weak long term evidence.
Researchers warned governments risk making major policy decisions based on exaggerated expectations about AI’s impact on efficiency and public services.
For now, however, the most immediate fear may not be whether AI changes the economy.
It may be whether the technology is already giving cybercriminals the power to launch faster, smarter and far more devastating attacks than ever before.