Ledger / May 2024
LLMjacking: stolen cloud credentials used to hijack hosted Claude and other LLM services
Confirmed
Sysdig, which investigated the intrusions and named LLMjacking, states what the stolen access was used for; its own report is cited.
- Model
- Claude, AWS BedrockAnthropic · closed; Amazon · closed
- Agent harness
- Nonechat or tool use only
- Attributed country
- Not attributedthe source names no country
- Sectors hit
- enterprise broad, ai platformsCloud accounts with exposed credentials (Laravel CVE-2021-3129 entry), victims billed up…
Attackers used cloud credentials stolen through a vulnerable Laravel system to abuse hosted LLM services on victims' accounts, including Claude on AWS Bedrock, and resold the access through reverse proxies.
“Attackers reportedly exploited stolen cloud credentials obtained through a vulnerable Laravel system (CVE-2021-3129) to allegedly abuse AI cloud services, including Anthropic’s Claude and AWS Bedrock, in a scheme referred to as “LLMjacking.”AI Incident Database
Sources
- LLMjacking: stolen cloud credentials used to hijack hosted Claude and other LLM servicesAI Incident Database · May 2024
- LLMjacking: stolen cloud credentials used in new AI attackSysdig · May 2024
- Date
- May 2024
- Actor
- Unnamed LLMjacking operators (criminal), as attributed by AI Incident Database
- Targets
- Cloud accounts with exposed credentials (Laravel CVE-2021-3129 entry), victims billed up to $100,000 per day
- What the AI did
- LLMjacking
- Model as named
- Claude, AWS Bedrock
- Model basis
- A researcher names it
- Autonomy
- Assisted a human
- Outcome
- Victims faced LLM consumption costs Sysdig estimated at up to $100,000 a day.