McDonald's McHire AI recruitment platform exposed 64 million applicants' data via default login and API flaw
Security researchers discovered that McDonald's McHire AI hiring platform used a default admin login of "123456" and contained an API vulnerability that could expose personal data of up to 64 million job applicants. McDonald's and Paradox.ai, the platform's developer, disabled the default credentials and patched the flaw after disclosure, stating that only five records were viewed by the researchers.
What happened
Security researchers Ian Carroll and Sam Curry discovered in late June 2025 that McDonald's McHire platform, which uses Paradox.ai's chatbot Olivia to screen job candidates, had a default admin login with username and password both set to "123456".
The researchers found that the administrative interface contained an insecure direct object reference (IDOR) flaw in an internal API that allowed access to applicant data by decrementing ID values, exposing names, emails, phone numbers, chat histories, shift preferences, personality test results, and authentication tokens.
Reported impact
The security flaws potentially exposed sensitive personal information of as many as 64 million job applicants, including full applicant profiles and chat logs.
Experts noted the vulnerability created risks for phishing, impersonation, and social engineering attacks using the exposed applicant data and tokens.
Index analysis
Aditi Gupta, senior manager at Black Duck Consulting, stated that "even sophisticated AI systems can be compromised by elementary security oversights" and that "the rush to deploy new technology must not compromise basic security principles."
Evan Dornbush of Desired Effect noted the incident exemplifies deploying technology without understanding how it can be exploited by untrusted users, and emphasized that organizations must invest in understanding and mitigating pre-emergent threats with AI systems handling sensitive data.
Governance controls
McDonald's and Paradox.ai disabled the default login credentials and fixed the vulnerable API endpoint after the researchers disclosed the flaws on June 30, 2025.
Paradox.ai stated that only five candidate records were accessed during the security review and that no data was leaked publicly.
Classification
Sources and evidence
Sources not yet published
This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.