1. News
  2. Alerts
  3. Security · AI bills
AlertSecurity · AI bills2 min read

Hackers are running up million-dollar AI bills on company accounts

Google says criminals are stealing company AI logins, reselling access, and leaving the victims with bills that reached $1 million at some firms.

By the Dashline deskSource: PYMNTS
Hackers are running up million-dollar AI bills on company accounts
Photo: Gryffindor, Public domain
Act this week
What happenedGoogle's threat team reported that criminals steal company AI credentials, resell access to premium AI models, and leave the victims with the usage charges.
Who is affectedCompanies with usage-billed cloud AI accounts, including firms in financial services, retail, technology and transportation across several continents.
What to do
  1. 1Find every AI key, service account and login your company uses, and remove any that are unused.
  2. 2Set hard spending limits and usage alerts on each AI account.
  3. 3Keep keys out of shared code folders and rotate them on a regular schedule.
  4. 4Ask finance to review cloud AI charges weekly and flag sudden spikes right away.

Criminals are breaking into company AI accounts, selling access to the paid AI tools, and leaving the business to pay the bill. Google's threat intelligence team reported the scheme on Monday, Nov. 2. It is called LLMjacking. At some companies, fake AI use topped $100,000 in a single day.

What happened

Google's team documented cases where total charges reached $1 million before access was shut off. The victims were in financial services, retail, technology and transportation, on several continents. The core problem is simple. A cloud company will bill an account for any request that carries a valid login. Nobody in finance has to approve the spending.

How the scheme works

Premium AI models charge by how much text they process. With a stolen login, an attacker can send huge numbers of requests and the account owner pays for all of them. Some criminals run what researchers call LLM Mixers. These services collect stolen logins and resell AI access at a discount.

One service, Jet Night, advertised unlimited access to top AI models for about $60 a month. Google linked it to an attacker who made more than 500,000 requests in under eight hours. That passed nearly $1 million in costs to one victim.

The break-ins are not exotic. Google points to exposed access keys, compromised service accounts, stolen identity tokens and logins left inside software code folders. In one case, attackers signed hacked accounts up for pay-as-you-go plans and picked premium models. They also created and deleted short-lived accounts to get around limits.

Who it affects

The risk falls on companies that pay for cloud AI by usage. If your bill rises with every request, a stolen login can raise it very fast. Businesses that do not run usage-billed AI accounts are not the target described in the report.

What to do

The alert box above lists the steps. They come down to knowing which AI logins exist, capping what they can spend, protecting the keys, and having finance look at the charges every week.

What happens next

Google expects the market for stolen AI access to grow as premium models and automated AI tasks spread. The report does not say how many companies have been hit. It also does not say whether cloud providers have reversed any of these charges, so a business should not assume it would get the money back.

The larger lesson is for finance teams. Spending approval used to mean a purchase order. With AI, it also means deciding which accounts and software keys are allowed to spend money, and how much.

The takeaway

If your company pays for AI by usage, set spending caps and alerts now, before a stolen login does the spending for you.

More articles

You should really read these.

Next step

Let’s talk