AI Beat

GuideFor your jobPolicy

AI for HR and recruiters: a practical guide

AI now writes job ads, screens CVs and schedules interviews. Here is what the tools do, why bias is the biggest legal risk, and which hiring laws apply in the US, the UK and the EU.

Two people shaking hands at a meeting
Photo: Mina Rad on Unsplash

In short

  • About four in ten organisations use AI in HR, and six in ten large ones, according to SHRM's 2026 survey. Recruiting is the most common use.
  • Bias is the core risk. In a University of Washington study, AI models preferred CVs with white-associated names 85% of the time and Black-associated names only 9%.
  • The employer stays responsible. A US court has let a discrimination case against Workday go ahead on the theory that a software vendor can act as the employer's agent.
  • Hiring is one of the most regulated uses of AI: New York City requires bias audits, Illinois bans discriminatory AI in employment, and the EU treats recruitment AI as high-risk from December 2027.
  • The EU has already banned emotion recognition at work, including in video interviews, since February 2025.
  • Use AI to save time on writing, scheduling and sourcing, keep people in charge of decisions, test for bias and tell candidates when AI is involved.

Hiring is where AI meets people’s livelihoods, which is why it is both one of the most useful and one of the most regulated uses of the technology. Recruiters are flooded with applications, many of them written with AI, and the software they use now ranks, screens and messages candidates on their behalf. This guide covers what the tools do, where the legal risk sits, and how to use them fairly.

Who uses AI in HR

SHRM’s 2026 survey of 1,908 HR professionals found that 39% of organisations use AI in HR, rising to 60% among those with 5,000 or more employees. Recruiting is the most common use. Most users report efficiency gains, but just under half of organisations have a formal AI policy. More strikingly, 57% of HR staff in states with AI employment laws didn’t know those laws existed.

The tools

Tool Made by What the AI does
Hiring Assistant LinkedIn Finds and contacts candidates, and drafts outreach messages
Workday Recruiting, HiredScore and Paradox Workday Matches and screens applicants; a chat agent handles frontline hiring and scheduling
SAP SuccessFactors with Joule SAP AI assistant and agents across HR processes, including recruiting
Greenhouse Real Talent Greenhouse Flags fraudulent or spam applications, verifies identity and matches talent without automating the decision
HireVue HireVue Video and voice interviews, and skills assessments
ChatGPT, Claude, Copilot, Gemini Various Job ads, interview questions, policies and employee communications

The safest uses are the ones that don’t decide anything about a person: writing a clearer job ad, drafting interview questions, summarising policy changes or scheduling. The riskiest are the ones that rank, score or reject candidates.

Why bias is the biggest risk

AI learns from past data, and past hiring was not neutral. When researchers test it, the bias shows up clearly. In a University of Washington study, three AI models compared more than 550 real CVs millions of times, with only the names changed.

Same CV, different name, different result

How often AI models preferred a CV with a name associated with each group, when the CVs were otherwise identical

  1. White-associated names85%
  2. Black-associated names9%
  3. Male-associated names52%
  4. Female-associated names11%

Source: Wilson and Caliskan, University of Washington, 2024

The rest of the comparisons showed no clear preference. The UK’s data regulator found similar problems in practice: its 2024 audits of AI recruitment tools found tools inferring gender and ethnicity from names, and filters that could screen out people with protected characteristics. It made almost 300 recommendations.

Courts are treating this as the employer’s problem, and increasingly the vendor’s too:

  • In 2023 the US Equal Employment Opportunity Commission settled with iTutorGroup for $365,000. The company’s software had automatically rejected women aged 55 or over and men aged 60 or over.
  • In Mobley v. Workday, a federal court in California has allowed a job seeker’s discrimination claims against Workday to proceed, on the theory that a software vendor can be liable as the employer’s agent. In 2025 the court let applicants aged 40 and over join as a collective. The case is ongoing.

What the law says

United States. No federal AI hiring law exists, but anti-discrimination law applies in full. States and cities are adding rules:

  • New York City has required independent bias audits and candidate notices for automated hiring tools since July 2023. A state audit in December 2025 found enforcement weak: the city had received only two complaints.
  • Illinois has made discriminatory use of AI in employment decisions, and failing to tell workers about it, a civil rights violation since 1 January 2026. Using ZIP codes as a stand-in for protected characteristics is banned.
  • California civil rights regulations on automated decision systems in employment have applied since October 2025. Employers must keep the related records for four years.
  • Colorado replaced its 2024 AI Act in 2026 with a narrower law on automated decisions, which takes effect on 1 January 2027 and focuses on notice, explanations and human review.

European Union. The AI Act lists AI used for recruitment, selection, promotion, dismissal, task allocation and monitoring of workers as high-risk. After a 2026 amendment, those obligations apply from 2 December 2027. One rule already applies: since February 2025 it has been illegal to use AI to recognise employees’ or candidates’ emotions, except for medical or safety reasons.

United Kingdom. The Data (Use and Access) Act 2025 relaxes the rules on fully automated decisions but keeps safeguards: people must be told, be able to make their case, get a human to look at the decision and challenge it. Equality law still applies to any tool you use.

How to use AI fairly

  1. Keep a human in the decision. Let AI draft, sort and summarise; have a person decide who is rejected, interviewed or hired.
  2. Ask vendors for evidence. Request bias audit results, what data the model uses and how it is tested. If they won’t say, don’t buy.
  3. Test your own results. Compare selection rates across groups regularly, as NYC’s law requires.
  4. Tell candidates. Say when AI is used and how they can ask for a human review or an adjustment for a disability.
  5. Don’t use emotion or personality reading from video or voice. In the EU it’s banned at work, and the science behind it is weak everywhere.
  6. Protect employee data. Keep personal data out of consumer chatbots; use tools under contract.

See how AI is changing hiring in our Job search and AI Demand Index, or browse the other guides.

Comments

What do you think? No account needed. Please be kind; comments that attack people are removed.

We store only your name and comment. See our privacy policy.

    Sources

    1. SHRM: State of AI in HR 2026 (March 2026)
    2. SHRM: AI in HR in 2026, from hype to measured, human-centred impact
    3. University of Washington: AI tools show biases in ranking job applicants' names (October 2024)
    4. Wilson and Caliskan: Gender, race and intersectional bias in resume screening via language model retrieval (arXiv)
    5. Civil Rights Litigation Clearinghouse: Mobley v. Workday
    6. US EEOC: iTutorGroup to pay $365,000 to settle EEOC discriminatory hiring suit (September 2023)
    7. New York State Comptroller: Enforcement of Local Law 144, automated employment decision tools (December 2025)
    8. Illinois General Assembly: HB 3773 (Public Act 103-0804)
    9. Colorado General Assembly: SB 26-189, automated decision-making technology
    10. California Civil Rights Council: Automated-decision systems regulations (June 2025)
    11. EU AI Act, Annex III: high-risk AI systems
    12. EU AI Act, Article 5: prohibited AI practices
    13. European Commission: AI Omnibus enters into force (July 2026)
    14. UK ICO: AI tools in recruitment, audit outcomes report (November 2024)
    15. UK ICO: The Data (Use and Access) Act 2025, what it means for organisations
    16. LinkedIn: Hiring Assistant is now globally available (September 2025)
    17. Workday: Workday completes acquisition of Paradox (October 2025)
    18. SAP: New SAP SuccessFactors innovations (October 2025)
    19. Greenhouse: Real Talent
    20. HireVue: HireVue acquires Hireguide technology (March 2026)

    More guides