Ledger / May 2026
Attacker uses an LLM agent after a marimo notebook exploit to exfiltrate an internal database
Strong
Sysdig investigated; its basis is the recorded command stream measured against 'the four signatures of agent-driven execution'. Behavioural inference.
- Model
- Not namedAI agents (model not named)
- Agent harness
- Nonechat or tool use only
- Attributed country
- Not attributedthe source names no country
- Sectors hit
- technologyA victim running a vulnerable marimo Python notebook (internal PostgreSQL database…
After exploiting a marimo notebook flaw, an attacker's LLM agent moved through four pivots to an internal database; Sysdig says the command stream showed live agent execution rather than a prebuilt script.
“On May 10, 2026, the Sysdig Threat Research Team (TRT) observed an intrusion driven by a large language model (LLM) agent in its post-exploitation phase.”Sysdig
Sources
- Date
- May 2026
- Actor
- Unidentified attacker (unknown), as attributed by Sysdig
- Targets
- A victim running a vulnerable marimo Python notebook (internal PostgreSQL database exfiltrated)
- What the AI did
- Agentic intrusion, Exploit dev
- Model as named
- AI agents (model not named)
- Model basis
- A researcher names it
- Autonomy
- Semi-autonomous
- Outcome
- An internal database was exfiltrated within four pivots of the initial exploit.