PreviewYou're previewing CHRO & People Compass. Sign in to use it.Sign in
← All lessons
L4 · HR agents and employee AI assistants9 min read

Employee data, privacy and security in HR AI

HR holds some of the most sensitive data in the company: identity, pay, health, performance, family, background checks. AI tools increase both the amount of data in motion and the number of places it can leak.

Four risks

  1. Employees pasting data into unapproved tools. In 2023 Samsung restricted staff use of generative AI tools after engineers uploaded sensitive information, including source code and meeting notes, to ChatGPT; it warned that violations could lead to dismissal.
  2. Weak security in HR-tech platforms. In 2025 security researchers found that McHire, the AI-chatbot hiring platform used by most McDonald's franchisees and run by Paradox.ai, had an administrator interface that accepted the default credentials "123456" plus an access flaw, which together could expose personal data from more than 64 million job applications. Paradox fixed the issues after being notified.
  3. Over-broad access. AI assistants that can search everything may show managers data they should not see.
  4. Vendors training on your data. Candidate and employee data used to improve a vendor's general model, without consent.

Controls that work

  • Data classes: public, internal, confidential, personal/sensitive — and which tools may use each.
  • Approved tools only for personal data, with contracts covering training rights, retention, location and breach notice.
  • Least-privilege access and logs for AI assistants and agents.
  • Security review as a launch gate for every HR AI tool, including vendor penetration-test evidence and basic hygiene like no default passwords.
  • Minimize: collect and keep only what the use needs.
  • Banned uses: for example emotion recognition and covert monitoring.

Tell people

Candidates and employees should know what data AI uses about them, why, and how to ask questions — in many places this is also a legal duty.

What this means for the CHRO

Put HR data rules at the center of your AI policy and your HR-tech procurement. Partner with the CISO: they bring the security review, you bring the data map and the people impact.

One leaked applicant database can undo years of employer-brand work — make security review a non-negotiable launch gate.

Ask for a list of every HR AI tool with the data it uses, where that data is stored, who can access it and whether the vendor can train on it. Close gaps in contracts at the next renewal, and make sure candidates and employees are told what AI does with their data.

Real example

In July 2025 security researchers Ian Carroll and Sam Curry reported that McHire, the AI-chatbot hiring platform used by 90% of McDonald's franchisees and built by Paradox.ai, had an administration interface that accepted the default credentials 123456:123456 and an access-control flaw that together allowed retrieval of personal data from more than 64 million applicants; the researchers reported the issues, which were fixed. Source: Ian Carroll — Would you like an IDOR with that? Leaking 64 million McDonald's job applications (Jul 2025) ↗

Use it

Quick check · 3 questions

  1. 1.Which control would have prevented the McHire default-password issue?

  2. 2.Why did Samsung restrict generative AI tools in 2023?

  3. 3.Which data should go only into approved tools with strong contracts?

0 of 3 answered

Sources

Last reviewed 2026-10-01.