HISPI Project Cerebellum HISPI PROJECT CEREBELLUM

Alleged LLMjacking Targets AI Cloud Services with Stolen Credentials

May 6, 2024

Stolen cloud credentials, possibly obtained via the vulnerable Laravel system (CVE-2021-3129), are suspected to have been exploited in a scheme known as 'LLMjacking.' This alleged attack targeted AI cloud services such as Anthropic’s Claude and AWS Bedrock. The attackers reportedly monetized access through reverse proxies, potentially causing daily costs of up to $100,000 for victims. Additionally, they are said to have bypassed sanctions, enabled LLM models, and evolved techniques to evade detection and logging—highlighting the need for safe and secure AI practices.

For those interested in shaping responsible AI governance and ensuring the harm prevention measures outlined by Project Cerebellum, JOIN US or explore how this incident maps to HISPI Project Cerebellum TAIM (Govern).

Matched TAIM controls

Suggested mapping from embedding similarity (not a formal assessment). Browse all TAIM controls

Alleged deployer
llmjacking-attackers-exploiting-laravel, entities-engaging-in-russian-sanctions-evasion
Alleged developer
oai-reverse-proxy-tool-creators, llmjacking-reverse-proxy-tool-creators
Alleged harmed parties
laravel-users, laravel-cve-2021-3129-users, cloud-llm-users, cloud-llm-service-providers

Source

Data from the AI Incident Database (AIID). Cite this incident: https://incidentdatabase.ai/cite/898

Data source

Incident data is from the AI Incident Database (AIID).

When citing the database as a whole, please use:

McGregor, S. (2021) Preventing Repeated Real World AI Failures by Cataloging Incidents: The AI Incident Database. In Proceedings of the Thirty-Third Annual Conference on Innovative Applications of Artificial Intelligence (IAAI-21). Virtual Conference.

Pre-print on arXiv · Database snapshots & citation guide

We use weekly snapshots of the AIID for stable reference. For the official suggested citation of a specific incident, use the “Cite this incident” link on each incident page.