Disclosed vulnerability
McDonald's McHire AI hiring chatbot exposed 64 million job applications via a 123456 test account and an IDOR
AIR-2025-0056 AIR record · 2025-07-mcdonalds-mchire-paradox-exposure
From the supporting source
name, address, phone number, email address, candidacy state, and an auth token to log into the consumer UIsecurityweek.com
The record
Disclosed
2025-07-11
Occurred
2025-06-30
Vendor
Paradox.ai / McDonald's
Product
McHire (Paradox.ai Olivia hiring chatbot)
Risk domain
Privacy › Sensitive data exposure
System type
Generative model
Who acted
AI was the target — default credentials plus an IDOR exposed 64M applications
Attack surface
Consumer assistant
Autonomy level
Assistant
Initial vector
Default test credentials plus insecure direct object reference
Impact type
Data exposure
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
64 million job applications - names, addresses, phone numbers, email addresses, candidacy state and consumer-UI auth tokens
Capability abused
Defense & remediation
Recorded defense
none effective - the 2019 test account password was never rotated
Guardrail kinds
None
Guardrail outcome
Not applicable
Fix shipped
both flaws confirmed resolved 2025-07-01
Taxonomy & references
Curator notes
Found by researchers Ian Carroll and Sam Curry; reported 2025-06-30, fixed 2025-07-01, published 2025-07-11 (Wired had the original story). impact_realized false: Paradox.ai says only five applicants' chat interactions were accessed and no candidate information was shared online. Included because the harm channel is an AI agent holding conversational PII at scale and the coverage was journalism-led rather than CVE-led.
Supporting sources
Cite as
Enkrypt AI. "McDonald's McHire AI hiring chatbot exposed 64 million job applications via a 123456 test account and an IDOR". AIR, AIR-2025-0056. https://agentincident.vercel.app/i/AIR-2025-0056/ (accessed 2026-10-19).