In the wild
Caught in the wild: protocol-aware attack traffic hitting exposed Clawdbot/OpenClaw gateways within minutes of a honeypot going live
AIR-2026-0021 AIR record · 2026-01-pillar-clawdbot-gateway-attack-traffic
From the supporting source
Every captured RPC method the attackers probed has a real handlerpillar.security
The record
Disclosed
2026-01-29
Vendor
OpenClaw
Product
Clawdbot / OpenClaw gateway (pre-2026-01-26 versions)
Risk domain
Access control and permissions › Credential theft
Also touches
Privilege escalation, Data exfiltration channels
System type
Generative model
Who acted
AI was the target — attack traffic aimed at exposed agent gateways
Attack surface
Autonomous operations
Autonomy level
Semi-autonomous
Initial vector
Direct attack on exposed agent gateway
Impact type
Credential theft, Data exfiltration, Remote control
Confidence
High
Submitted by
EAI Internal
Crosswalk
- OWASP Agentic Top 10 (2026)
- ASI03 – Identity and Privilege Abuse
- MITRE ATLAS
- AML.T0012 – Valid Accounts
- NIST AI RMF
- MEASURE 2.7
- EU AI Act
- Article 13
- ISO/IEC
- TR 24028; 42001; 23894
- AIUC-1
- B. Security
- Scope
- tool failure
- Mitigation
- Use secure credential storage; Implement short-lived tokens.
- Taxonomy
- Enkrypt AI agent risk taxonomy · Black-Box Red Teaming of Agentic AI: A Taxonomy-Driven Framework for Automated Risk Discovery
Blast radius
internet-exposed Clawdbot gateways on unpatched versions
Capability abused
Defense & remediation
Recorded defense
gateway authentication (patched 2026-01-13 to 2026-01-26)
Guardrail kinds
Authentication
Guardrail outcome
Silent
Fix shipped
upstream patches 2026-01-13 to 2026-01-26
Taxonomy & references
Curator notes
Pillar Security honeypot study. impact_realized=false because the observed traffic hit a honeypot, not confirmed real victims - but the traffic itself is real and ongoing. Attackers bypassed the AI layer entirely and spoke the WebSocket control protocol directly, with a three-phase playbook (recon, enumeration, exploitation); 67% of traffic from a single source IP (144.31.99.48). Successful compromise yields Anthropic/OpenAI API keys, gateway credentials, Telegram/Discord/Slack channel tokens, conversation history and a multi-node infrastructure map. Complements 2026-01-clawdbot-exposed-control-interfaces (exposure census) with the attacker-side view. verified_quote is a verbatim clause.
Supporting sources
Cite as
Enkrypt AI. "Caught in the wild: protocol-aware attack traffic hitting exposed Clawdbot/OpenClaw gateways within minutes of a honeypot going live". AIR, AIR-2026-0021. https://agentincident.vercel.app/i/AIR-2026-0021/ (accessed 2026-10-19).