Attackers are taking control of access to corporate large language models (LLMs) and reselling it, leaving victims with cloud bills reaching $1 million, according to reporting that cites Mandiant.
AI Security · Cloud Costs
Hijacked AI access is leaving victims with cloud bills of up to $1 million
Criminals take control of corporate large language model services, resell that access, and the companies whose accounts were compromised pay for the computing consumed.
$1M
Reported cloud bill tied to unauthorized LLM use
LLMHijacking
The name given to this pattern of access being stolen and resold
Scale of the reported bill
Each block represents $100,000 of the reported $1,000,000 total.
The figure is presented as a single reported amount. It is not clear whether it reflects one incident or a range of cases.
How the harm flows to the victim
1. Criminals gain access to an AI service run through a cloud platform
→
2. That access is sold to others
→
3. The compromised organization is billed for tokens and compute used
→
4. Charges can grow rapidly before detection
Open questions
The material does not yet explain the full picture
How the access was obtained
Which organizations were affected, and how many
How the resale operation was run
Whether the $1 million figure is one incident or a range of cases
Monitor usage and set alerts
Track AI service consumption and set spending alerts so unusual growth is caught early.
Protect credentials and access paths
Secure the credentials and routes that connect applications and tools to these models.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article...