Research demo
LLM data exfiltration via messaging-app URL previews: OpenClaw's default Telegram configuration leaks data with no click
AIR-2026-0146 AIR record · 2026-07-openclaw-url-preview-exfiltration
From the supporting source
OpenClaw (the current trending agentic system) is vulnerable to this attack when using the default Telegram configuration; an example secure configuration is provided at the end of the article.promptarmor.com
The record
Disclosed
2026-07
Vendor
OpenClaw
Product
OpenClaw (default Telegram configuration); any LLM agent replying through Telegram/Slack
Risk domain
Privacy › Data exfiltration channels
Also touches
Confused deputy
System type
Generative model
Who acted
Agent acted — the agent emitted links whose previews leaked the data with no click
Attack surface
Autonomous operations
Autonomy level
Semi-autonomous
Initial vector
Indirect prompt injection
Impact type
Data exfiltration
Confidence
High
Submitted by
EAI Internal
Crosswalk
- OWASP Agentic Top 10 (2026)
- ASI02 – Tool Misuse and Exploitation
- MITRE ATLAS
- AML.T0024 – Exfiltration via AI Inference API
- NIST AI RMF
- MAP 4.2
- EU AI Act
- Article 10
- ISO/IEC
- TR 24028; 23894
- AIUC-1
- A. Data & Privacy
- Scope
- tool misuse
- Mitigation
- Monitor network traffic for anomalous patterns; Restrict egress destinations.
- Taxonomy
- Enkrypt AI agent risk taxonomy · Black-Box Red Teaming of Agentic AI: A Taxonomy-Driven Framework for Automated Risk Discovery
Blast radius
OpenClaw users on the default Telegram config, and LLM agents that answer in link-previewing chat apps generally
Capability abused
Defense & remediation
Recorded defense
none by default
Guardrail kinds
None
Guardrail outcome
Not applicable
Fix shipped
mitigation only: config change to disable link previews
Taxonomy & references
Curator notes
PROVENANCE CAVEAT: no publication date on the page; 2026-07 in the incident_id is provisional and date_disclosed is left null. PromptArmor Threat Intel Team. Same primitive Johann Rehberger described for Slack unfurling in 2024 (see 2025-06-anthropic-slack-mcp-unfurl), generalized here to any agent that replies into a preview-generating chat app. | date_disclosed is recorded to the month only; no source establishes the exact day.
Supporting sources
Cite as
Enkrypt AI. "LLM data exfiltration via messaging-app URL previews: OpenClaw's default Telegram configuration leaks data with no click". AIR, AIR-2026-0146. https://agentincident.vercel.app/i/AIR-2026-0146/ (accessed 2026-10-19).