The era of cryptojacking now has a more profitable successor in LLMjacking. As enterprise-grade AI inference costs climb, attackers have pivoted from hijacking machines to mine cryptocurrency toward draining cloud billing accounts for unauthorized access to premium models. Surging demand for those models has fueled an underground broker network that trades stolen credentials and API keys like liquid assets, leaving victim companies to cover the tab for illicit usage.
Quick Take
- What happened: Threat actors are systematically hijacking cloud billing accounts and developer platforms to funnel unauthorized, high-volume AI inference costs to corporate victims.
- Why it matters now: Premium AI access requires vast compute. Attackers exploit cloud credentials to bypass traditional malware defenses, racking up tens to hundreds of thousands of dollars in illicit Gemini or OpenAI usage in mere days.
- Who is most affected: Cloud security architects, FinOps teams, and enterprise developers relying on expensive models and AI coding agents like Devin and Cursor.
- The under-reported angle: Hackers are operating like illicit SaaS vendors, offering up to 97% discounts on premium AI access complete with replacement guarantees to counter account suspensions.
What happened
Google’s Threat Intelligence Group and Wiz have flagged a sharp rise in stolen AI access on dark-web markets. During the stretch they call “Black October,” attackers used phishing and infostealers to harvest cloud credentials from AI developer platforms. Once inside GCP or AWS environments, they link projects across tenants or alter IAM policies so expensive APIs such as Gemini or Claude run from attacker-controlled systems while the victim’s account pays the bill. Prices for these brokered accounts reportedly doubled through 2026.
Why it matters now
The shift tracks directly with the rising cost of high-end model inference. Rather than simply stealing data, attackers now target the compute budgets themselves. Because the activity uses legitimate cloud APIs and valid identities, it evades malware-based detection. Bills can balloon within a 24- to 48-hour window, and because the workloads often run in linked or dormant projects, victims frequently have no visibility into the prompts or outputs involved.
Who is most affected
The attack sits at the intersection of security and spend, affecting anyone who controls cloud keys or AI platform access.
| Stakeholder | Impact | Why |
|---|---|---|
| Cloud Security & SOC Teams | Blind spots in detection | LLMjacking relies on legitimate API usage and IAM policy shifts, generating no traditional malware signatures. |
| FinOps & Cloud Admins | Rapid, devastating cost spikes | Attackers exploit high-quota API limits, generating staggering bills in hours before budgets cap the damage. |
| AI Developers | Targeted credential theft | Local .env files, browser stores, and GitHub tokens are actively hunted to access tools like Cursor Pro and Devin. |
The Broker Economy of Stolen Intelligence
The real danger lies in the organized resale layer that has formed around the thefts. Vendors now run what amounts to an “AI-as-a-Service” black market, offering discounted access to models from OpenAI, Anthropic, and Google. To keep customers despite provider suspensions, many brokers include replacement guarantees and basic support—effectively turning the operation into a reliable supply chain that lowers the cost for other criminals to run phishing, extortion, or model development on someone else’s tab.
About the analysis
Researched and written with AI assistance under the editorial responsibility of the i10X News team. The title image is an AI-generated illustration. Read our AI Use Policy. Spotted an error? Request a correction.
This independent i10x analysis synthesizes threat intelligence and market data to provide CTOs, security leaders, and developers with an operational understanding of emerging AI infrastructure risks.
- From Phishing to LLMjacking: Threat Actors Targeting Cloud Billing Accounts for LLM Usage | Wiz Blog
- What is LLM Jacking? | Wiz
- LLM jacking fuels underground market for AI access
- Google warns of surge in AI account theft and LLM-jacking attacks
- LLM-jacking emerges as a blowing form of AI-enabled cybercrime, with underground prices doubled over 2026 | Digital Watch Observatory
- Cloud Billing Hijacking as LLMjacking
- Black October for AI developer platforms
i10x Perspective
LLMjacking marks a clear step in how threats around AI are evolving: the models themselves have become the most tradable illicit commodity. We’ve seen security teams begin to treat inference budgets with the same rigor once reserved for data protection. When billing permissions are not locked down as tightly as any other high-value control, the infrastructure is already set up to underwrite the next round of black-market activity.
