In the wild
Realized harm
Attackers reached an AI call-centre platform's management dashboard and took 10.2 million consumer, operator and AI-agent conversations
AIR-2024-0043 AIR record · 2024-10-middle-east-ai-call-centre-breach
A real party was affected. This incident caused actual consequences, not a demonstrated capability.
From the supporting source
According to cybersecurity firm Resecurity, the breach involved unauthorized access to the platform's management dashboard, allowing attackers to collect over 10.2 million interactions between consumers, operators and AI agents.resecurity.com
The record
Disclosed
2024-10-08
Occurred
2024-10
Vendor
undisclosed
Product
unnamed Middle Eastern AI-powered cloud call-centre platform
Risk domain
Privacy › Sensitive data exposure
Also touches
Data exfiltration channels
System type
Generative model
Who acted
AI was the target — the management dashboard was compromised and the transcript store taken
Attack surface
Multi-agent system
Autonomy level
Semi-autonomous
Initial vector
Management dashboard compromise
Impact type
Data exfiltration, Privacy violation
Confidence
High
Submitted by
EAI Internal
Crosswalk
- OWASP Agentic Top 10 (2026)
- ASI06 – Memory & Context Poisoning
- MITRE ATLAS
- AML.T0057 – LLM Data Leakage
- NIST AI RMF
- MEASURE 2.10
- EU AI Act
- Article 10
- ISO/IEC
- TR 24028; 23894
- AIUC-1
- A. Data & Privacy
- Scope
- tool misuse
- Mitigation
- Use data anonymization; Implement log sanitization to scrub PII.
- Taxonomy
- Enkrypt AI agent risk taxonomy · Black-Box Red Teaming of Agentic AI: A Taxonomy-Driven Framework for Automated Risk Discovery
Blast radius
over 10.2 million interactions including national ID documents, across fintech and e-commerce clients; offered for sale on the dark web
Capability abused
Defense & remediation
Recorded defense
dashboard authentication
Guardrail kinds
Authentication
Guardrail outcome
Silent
Fix shipped
undisclosed
Taxonomy & references
Curator notes
Disclosed by Resecurity. The lesson is where the value concentrated: not the models but the transcript store an agent platform necessarily accumulates, reached through the orchestration control plane rather than through any model. Victim platform never named, so target_vendor is undisclosed. AI Incident Database incident 906.
Supporting sources
Cite as
Enkrypt AI. "Attackers reached an AI call-centre platform's management dashboard and took 10.2 million consumer, operator and AI-agent conversations". AIR, AIR-2024-0043. https://agentincident.vercel.app/i/AIR-2024-0043/ (accessed 2026-10-19).