AI resume screening is not banned in India, and no dedicated AI-hiring statute legalises it either. The DPDP Act, 2023 still covers candidate personal data. The employer or staffing agency remains liable for how that data is used. You still need purpose, consent, and a way to explain a shortlist to a client. This page is not legal advice.
This is an operational briefing for Indian staffing desks, not a legal opinion. Laws and rules change. For a live req, talk to your counsel. Nothing here creates an attorney–client relationship or a compliance certificate for any tool, including TeamOB - SafalHires.
Software does not absorb liability. If your name is on the submission, the decision is still yours.
There is no AI-hiring licence — and no AI-hiring ban
Recruiters ask “is this legal?” as if there were a gazette notification for match percentages. There is not. Ranking resumes you already collected for a live requirement is an ordinary use of hiring tools, the same family as a Boolean search on Naukri. What is not ordinary is pretending the model is a separate legal person, or using candidate data for a new purpose the person never agreed to.
So the useful question is not “is AI legal?” It is: are we processing personal data for recruitment, with a purpose we can state, and a human who can still override the rank?
DPDP Act 2023 and candidate personal data
Names, phones, emails, CVs, CTC, notice, and location on a resume are personal data. The DPDP Act, 2023 is built around purpose, consent, and security — not around a whitelist of algorithms. Practical handling of tools is in data privacy rules for AI hiring tools in India. For screening, the desk-level translation is: collect for hiring, rank for hiring, do not enrich from social graphs, keep identity masked until unlock if that is your process, and do not keep CVs forever “in case.”
The score should use resume text and the JD, not a side project in marketing or a scrape of the open web.
Candidates who applied, or whose data you hold through a job board relationship, still need a clear hiring purpose. Do not treat a forwarded PDF as a blank cheque for every future client.
Who inside the agency can unlock a profile, export a CV, or send to a client should be a logged decision, not a shared mailbox.
| Practice | Why it matters |
|---|---|
| Rank on CV + JD | Purpose-limited processing; explainable to a client |
| Mask until unlock | Less identity in the first pass; unlock is auditable |
| Human override | The agency still makes the decision; the model is a sort key |
| Auto-reject with no review | Harder to explain; easier to own a discriminatory pattern |
Explainability, client contracts, and who remains liable
Clients will ask why this profile and not that one. “The AI said 81%” is a weak answer in a dispute and a weak answer in a vendor review. Show match percentage, missing must-haves, and recruiter notes. The script is in how to explain AI screening decisions to a client. If the tool discriminates, liability does not evaporate into the vendor’s terms of service. That question is taken up in who is liable if an AI recruiting tool discriminates.
Not the neural net. The features: these skills matched, this must-have was missing so the score was capped at 55%, this recruiter confirmed notice. That is enough for most hiring managers and closer to what DPDP-era accountability looks like in practice.
Practical checks before you turn screening AI on
Write down the purpose: filling client reqs. Confirm how candidate consent is captured on your apply path and job-board flow. Confirm the vendor does not train a public model on your CVs without a contract you have read. Confirm a recruiter can override. Confirm you can delete or stop using a profile when the purpose ends. None of that is a court opinion. All of it is cheaper than discovering, after a client complaint, that nobody could say what the score used.
AI resume screening can be used lawfully in India the way any ATS is used: as a tool under an employer’s (or agency’s) duty of care. It is not a loophole. It is not illegal by default. It is only as defensible as your purpose, your consent story, and your ability to explain the shortlist without hiding behind the model.
Screening you can explain
TeamOB - SafalHires ranks on the JD, keeps identity masked until unlock, and leaves the hiring decision with your recruiters — so the shortlist is a process you can describe.
Free trial