In the wild
Realized harm
AkiraBot used OpenAI's API plus CAPTCHA bypass to autonomously spam 80,000+ website contact forms and chat widgets
AIR-2025-0021 AIR record · 2024-09-akirabot-llm-spam-at-scale
A real party was affected. This incident caused actual consequences, not a demonstrated capability.
From the supporting source
SentinelLABS assesses that AkiraBot has targeted more than 400,000 websites and successfully spammed at least 80,000 websites since September 2024.sentinelone.com
The record
Disclosed
2025-04-09
Occurred
2024-09-01
Vendor
OpenAI (abused)
Product
OpenAI chat API (abused) via the AkiraBot framework
Risk domain
Governance › Policy drift
System type
Generative model
Who acted
Agent acted — the bot browsed, solved CAPTCHAs and submitted 80,000 forms itself
Attack surface
Autonomous operations
Autonomy level
Fully autonomous
Initial vector
Misuse of llm api by operator
Impact type
Spam at scale, Abuse
Confidence
High
Submitted by
EAI Internal
Crosswalk
- OWASP Agentic Top 10 (2026)
- ASI01 – Agent Goal Hijack
- MITRE ATLAS
- AML.T0010 – AI Supply Chain Compromise
- NIST AI RMF
- GOVERN 1.5
- EU AI Act
- —
- ISO/IEC
- TR 24028; 23894
- AIUC-1
- E. Accountability
- Scope
- agent failure
- Mitigation
- Implement version pinning for models and prompts; Monitor for behavioral drift.
- Taxonomy
- Enkrypt AI agent risk taxonomy · Black-Box Red Teaming of Agentic AI: A Taxonomy-Driven Framework for Automated Risk Discovery
Blast radius
more than 400,000 domains targeted; at least 80,000 successfully spammed
Capability abused
Defense & remediation
Recorded defense
site CAPTCHAs and anti-spam filters
Guardrail kinds
Unknown
Guardrail outcome
Silent
Fix shipped
OpenAI disabled the abused API key after disclosure
Taxonomy & references
Curator notes
AI Incident Database incident 1365; SentinelLABS research. Included because the bot autonomously browsed, solved CAPTCHAs and submitted forms - not merely generated text.
Supporting sources
Cite as
Enkrypt AI. "AkiraBot used OpenAI's API plus CAPTCHA bypass to autonomously spam 80,000+ website contact forms and chat widgets". AIR, AIR-2025-0021. https://agentincident.vercel.app/i/AIR-2025-0021/ (accessed 2026-10-19).