Hackers are eyeing the AI "gold mine": stealing accounts and hijacking cloud computing power, AI resources are becoming a new business for cybercrime.

AI accounts and computing power are becoming the most sought-after commodities in the underground market for cybercrime. Hackers are eyeing large, expensive language models for use in ransomware, cyber warfare, and espionage.

In a recent interview with the Financial Times, John Hultquist, chief analyst at Google Threat Intelligence Group, stated that there has been a significant increase in hacking attacks targeting AI accounts and computing resources this year, with a new attack method known as "LLM-jacking" gaining scale in the cybercrime underworld.

Darknet marketplaces are selling access to AI models from companies like Anthropic, Google, and OpenAI at discounts of up to 97%. Meanwhile, criminal gangs and state-sponsored hacking groups are infiltrating corporate cloud servers, implanting their own AI models to "steal" computing power—a tactic strikingly similar to the previous hijacking of others' machines for cryptocurrency mining.

“What we’re seeing in the underground market is an economy that’s growing around access to AI,” said the analyst, who has 20 years of experience in cybersecurity.

Dark web discounts up to 97% have led to the buying and selling of AI accounts becoming a gray industry.

Researchers at Google Threat Intelligence Group have discovered that dark web marketplaces are selling access to AI models from companies such as Anthropic, Google, and OpenAI at discounts of up to 97%.

It's worth noting that the highest-tier subscriptions to ChatGPT and Claude can cost up to $200 per user per month.

More notably, because AI companies proactively monitor signs of abuse, some sellers have even launched "guaranteed access" services—promising to reissue new credentials free of charge if the initial account is banned. This mechanism makes the black market supply chain operate more stably.

Hijacking cloud servers to "freeload" computing power

Besides reselling accounts, another type of attack is more direct, Hultquist revealed:

Some criminal gangs and state-sponsored hacking groups are infiltrating corporate cloud hosting servers and deploying their AI models directly on the target systems—at the expense of the victims, while the computing power is used by the attackers.

This is exactly the same logic behind how hackers used to break into other people's machines to "mine cryptocurrencies" in the past.

The defending side faces a "cost mismatch" dilemma.

Hultquist points out that the danger of this attack pattern lies not only in the technical aspects, but also in the economic imbalances.

“Ultimately, these actions give them an economic or efficiency advantage—because they can acquire this computing power at a much lower cost, while we have to pay full price to defend it,” he said.

Anthropic's latest quarterly AI abuse report reveals that threat actors have been found attempting to use its Claude tools for malicious activities in more than 20 countries , including the United States, the United Kingdom, and Yemen.

Hultquist's attitude is blunt: "Every threat actor is using AI."

New risk window: The initial stage of AI deployment is the best time to "hide".

As more and more large enterprises choose to build their own servers to deploy customized AI models instead of renting computing power from cloud service providers, these internal systems will also become new targets of attack.

Hultquist warns: "If you're paying for computing power—which can be very expensive—then it becomes a significant potential resource for threat actors."

He also pointed out that the period when companies have just completed the deployment of AI infrastructure is precisely the best time for hackers to "infiltrate".

You might think that a sudden surge in computing power usage is perfectly normal, since you've just introduced a bunch of AI infrastructure. This gives attackers a real opportunity to hide in the noise.

He concluded by warning: "Anyone who thinks AI is just a passing fad and wants to wait for it to pass will one day find themselves drowning. They will encounter more accidents, more alarms, and more attacks than ever before. We must get our positions sorted out now."

Risk Warning and DisclaimerInvesting involves risk; please exercise caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Any investment decisions made based on this information are at your own risk.