Vinsys
toggle
close
    • blog
    • dpdp act employee ai usage hr guide
    blog image

    DPDP Act and Employee AI Usage: What HR Must Know

    Table of Content
    DPDP timeline: what applies now and what comes nextDoes the DPDP Act apply to employee data?Why employee AI usage creates DPDP risk7 Steps Action Plan for HRTraining employees on AI and data protectionWho should own AI and DPDP Compliance
    Share Now

    India DPDP Act holds organisations responsible for how personal data is processed. When employees use AI tools on personal data, the organisation must assess the purpose, lawful basis, safeguards and vendor arrangements. HR should set an AI usage policy, train staff and prepare for the phased deadlines.

     

    Key takeaways

     

    • Employers act as Data Fiduciaries when they decide why and how employee data is processed.
    • Employee and candidate data entered into AI tools creates privacy, security and compliance risk.
    • HR, legal, IT security and compliance should jointly set approved AI rules.
    • Review AI vendors, contracts, retention and security controls.
    • Train staff on data minimisation, safe prompting, output checks and incident reporting.
    • The main provisions commence on 13 May 2027, so start early.

     

    Introduction

     

    Employees now use generative AI to draft emails, summarise resumes, write performance notes and analyse complaints. When these tasks involve personal data and unapproved accounts, privacy and compliance risks follow. The Digital Personal Data Protection Act, 2023 (DPDP Act) and the DPDP Rules, 2025 commence in phases, so HR needs to prepare now.

     

    DPDP timeline: what applies now and what comes next

     

    Commencement notification G.S.R. 843(E), dated 13 November 2025, sets three phases.

     

    Phase

    Commencement

    What starts

    Phase 1

    13 Nov 2025

    Definitions and Data Protection Board provisions

    Phase 2

    13 Nov 2026

    Section 6(9) (Consent Managers) and Section 27(1)(d)

    Phase 3

    13 May 2027

    Main provisions: grounds for processing, notice, consent, fiduciary obligations, individual rights, penalties

     

    The Rules have their own schedule. Rules 1, 2 and 17 to 21 commenced on publication, Rule 4 on 13 November 2026, and Rules 3, 5 to 16, 22 and 23 on 13 May 2027. Phase 2 covers Consent Managers only, so employer duties on consent and employee data start in Phase 3. Treat the window before May 2027 as preparation time.

     

    Does the DPDP Act apply to employee data?

     

    Yes. Digital personal data of employees and candidates is covered, subject to the Act's scope and exemptions. HR handles names, contact details, Aadhaar and PAN, resumes, payroll and bank data, appraisals, leave and health information, and disciplinary records.

     

    The Act does not ban AI, but using AI does not remove the duty to run a lawful, safeguarded process. If a recruiter uploads resumes to an AI tool, assess the purpose, lawful ground, the tool's data handling and the safeguards.

     

    Who is who in an HR context

     

    DPDP term

    Meaning

    HR example

    Data Fiduciary

    Decides the purpose and means of processing

    Employer

    Data Principal

    The individual the data relates to

    Employee or applicant

    Data Processor

    Processes data on the Fiduciary's behalf

    Payroll or HRMS provider

     

    Under Section 8, the employer stays responsible for processing done by a Processor. The same test applies to an external AI service: look at its real role, contract and data handling. The Act has no separate "sensitive data" category, but financial, identity and health data deserve stronger safeguards.

     

    Lawful grounds for processing employee data

     

    Using an AI tool does not create a lawful ground.

     

    • Consent (Section 6): must be free, specific, informed, unconditional and unambiguous, and as easy to withdraw as to give. Given the power imbalance at work, use it for optional activities such as photo use, not as blanket authorisation.
    • Legitimate use (Section 7(i)): covers processing for employment, or for safeguarding the employer from loss or liability, such as protecting trade secrets. It is not unrestricted: a valid employment purpose does not justify uploading full employee records into a public AI tool.
    • Notice (Section 5): must state the data processed, the purpose and how to exercise rights, with further detail in the Rules. Check whether existing notices still describe AI-based processing.

     

    Three principles for HR data

     

    1. Purpose limitation: do not reuse payroll data for unrelated profiling.
    2. Data minimisation: share only what the task needs. An employee number is not anonymisation if the person is still identifiable.
    3. Retention and deletion: know how long prompts and uploads are kept, what deletion the provider offers and whether copies persist elsewhere.

     

    Employee rights and employer duties

     

    • Rights (Sections 11 to 14): access, correction and completion, erasure (not unconditional), grievance redressal and nomination. HR needs a process to handle requests.
    • Security (Section 8): reasonable safeguards, including role-based access, approved tools and company accounts, authentication, monitoring, vendor review and staff training.
    • Breach notification (Rule 7): inform affected individuals and the Data Protection Board without delay, with a detailed report to the Board within 72 hours unless extended. If someone uploads an employee database to an unapproved AI service, start incident response at once. Deleting the file does not remove reporting duties.
    • The Board: examines breaches and complaints, directs remedies and imposes penalties. Keep documented policies, training records, vendor assessments and incident logs.

     

    Why employee AI usage creates DPDP risk

     

    The Act does not ban workplace AI. Risk arises when personal data is processed without a clear purpose, lawful ground, safeguards or oversight.

     

    • Shadow AI: personal accounts and unapproved tools. Keep an approved list, offer an approval route and train staff to report it.
    • Personal data in prompts: a pasted appraisal, resume or complaint is personal data. Minimise identifiers or use synthetic examples.
    • Data leaving the organisation: check the provider's role, storage location, subcontractors and retention. Section 16 lets the Government restrict transfers to specified countries. Overseas processing is neither automatically banned nor automatically allowed.
    • Accuracy and fairness: AI can err or reflect bias. Never use an AI summary as the sole basis for hiring, termination, pay or promotion.
    • Retention: deleting a chat does not necessarily delete the data. Practices vary by provider and plan.

     

    Six scenarios HR should plan for

     

    Scenario

    What can go wrong

    Control

    Recruiter pastes resumes into a public AI tool

    Candidate data processed without basis, notice or safeguards

    Approved tools, anonymised profiles

    Manager summarises appraisal notes

    Confidential data exposed; biased or wrong summary

    Restricted access, manager review

    Support staff paste complaints

    Customer identifiers leave approved workflows

    Mask identifiers, role-based access

    HR drafts a disciplinary letter

    Confidential data exposed; unproven claims enter the record

    Templates with placeholders, HR or legal review

    Team uses an AI meeting assistant

    Unannounced capture of employee or customer data

    Approve the tool, disclose recording, set retention

    Employee leaves with work chats in a personal AI account

    Data stays outside company control

    Company accounts, AI step in offboarding

     

    7 Steps Action Plan for HR 

     

    HR works with legal, IT, security and procurement.

     

    1. Map data and AI tools. Inventory recruitment, HRMS, payroll, performance and AI tools, including unapproved ones. Record purpose, access, storage and sharing.
    2. Document the lawful ground for each use, including whether Section 7(i) applies, and update notices when AI changes how data is used.
    3. Apply security safeguards: role-based access, multi-factor authentication, upload limits, logging and access reviews. A tool approved for drafting may not be approved for payroll or health data.
    4. Check vendor contracts for the provider's role, purposes, storage, retention, security, subcontractors and breach reporting. A privacy policy alone is not enough.
    5. Set up breach response so employees know whom to tell and Rule 7 notifications can be met.
    6. Provide a grievance channel and handle access, correction and erasure requests with clear owners.
    7. Train employees by role on what not to enter, anonymisation, approved tools and incident reporting.

     

    Employee AI usage: do and don't

     

    Do: use approved tools and company accounts; anonymise and share only what is needed; verify outputs; keep human review on people decisions; report accidental exposure immediately.

     

    Don't: upload identifiable employee or candidate data, or payroll, medical and disciplinary records, to unapproved tools; use personal AI accounts for confidential work; bypass security controls; treat AI output as the sole basis for a decision; hide mistakes, because deleting a prompt may not remove what was processed.

     

    AI usage policy: template outline

     

    1. Purpose and scope, including contractors and temporary staff
    2. Approved and restricted tools, with an approval process and account rules
    3. Data classification, with example prompts
    4. Permitted and prohibited use cases
    5. Personal data protection: minimisation, anonymisation, retention
    6. Human review and accountability
    7. Vendor approval and security requirements
    8. Training and awareness
    9. Incident reporting and escalation
    10. Enforcement, audits and review

     

    Training employees on AI and data protection

     

    Training turns policy into practice. Treat it as an ongoing programme, not a one-off.

    • Build DPDP awareness and explain why approved tools and defined purposes matter.
    • Show what can and cannot be shared with realistic examples: IDs, CVs, payroll, medical and disciplinary records, customer data.
    • Teach safe prompting: remove identifiers, use fictional examples, share only what the task needs.
    • Make it role-based: recruiters on CV handling, HR and payroll on salary and ID data, managers on human review of AI insights, IT on tool management and incident response.
    • Use scenarios and measure results: completion rates, assessment scores, incidents and reporting awareness. Refresh quarterly and after incidents.

     

    Vinsys helps organisations build practical AI capabilities through role-based training that connects AI tools with real business workflows and responsible usage.

     

    Vendor and tool due diligence checklist

     

    Involve IT, legal, procurement and security before approving any AI platform for employee data. Check:

     

    • Data use: what it collects, and whether submitted data trains the provider's models.
    • Storage and deletion: locations, retention periods, admin settings, user deletion, backups, and return or deletion at contract end.
    • Security: multi-factor authentication, role-based access, encryption, activity logs, independent certifications.
    • Contract: the provider's Processor role, documented instructions, subcontractors, breach reporting, deletion duties.
    • Cross-border transfers: processing locations, overseas subprocessors, Section 16 and sector rules.
    • Oversight and support: stated limitations, room for human approval, a documented incident process.

     

    Incident response: personal data entered into an AI tool by mistake

     

    Employees should report at once and not try to fix it alone.

     

    1. Report through a defined channel: tool, account (personal or company), time, data type and files uploaded.
    2. Contain: stop further uploads, restrict the account, revoke integrations and request removal. Deletion is not proof that all copies are gone.
    3. Assess: whose data, whether identifiable, whether it includes financial, health or ID details, and whether the provider retains or trains on it.
    4. Involve legal and compliance to decide whether it is a reportable breach, including the 72-hour Board report.
    5. Document and notify affected individuals where required.
    6. Prevent a repeat: review tool approval, tighten controls and use the case in training.

     

    HR coordinates and supports communication, IT security leads containment, and legal leads the regulatory assessment.

     

    Readiness roadmap: now to May 2027

     

    These phases are a recommended schedule, not statutory dates.

     

    Timeline

    Focus

    Key activities

    Now to Dec 2026

    Visibility and ownership

    Inventory HR data and AI tools, find unapproved AI, assign owners

    Dec 2026 to Jan 2027

    Policies and governance

    AI policy, documented purposes and lawful grounds, notice review, vendor approval

    Feb to Mar 2027

    Security and operations

    Access controls, retention, incident workflows, rights requests, drills

    Apr to 12 May 2027

    Training and validation

    Role-based training, policy acknowledgement, vendor gap closure, readiness assessment

    13 May 2027 onwards

    Operational compliance

    Ongoing monitoring and review

     

    Track progress monthly through indicators such as tools inventoried, vendors vetted, employees trained, high-risk use cases reviewed and open gaps.

     

    Who should own AI and DPDP Compliance

     

    Responsibility stays with the organisation as Data Fiduciary, but the work is shared.

     

    • HR: employee data governance, AI policy, training, rights requests
    • Legal and Compliance: interpreting the Act, lawful grounds, notices, breach advice
    • IT and Security: tool inventory, access, retention, incident response
    • Procurement: vendor due diligence and contracts
    • Leadership: oversight and resourcing

     

    Form a cross-functional AI governance group. A Data Protection Officer is mandatory only for Significant Data Fiduciaries.

     

    Example: a mid-size company's rollout

     

    A 500-person firm finds recruiters, managers, finance and support staff using AI, some on personal accounts. It inventories usage, sets approved rules through a cross-functional group, reviews HR data processing and vendors, trains by role, and monitors training completion, approved-tool use, incidents and new tools. The lesson: blocking public tools is not enough. Understand real behaviour, create approved workflows, vet vendors and train people.

     

    Frequently Asked Questions

     

    1. Does the DPDP Act apply to employee data?

    Yes. It covers digital personal data of employees and candidates, subject to scope and exemptions. Employers generally act as Data Fiduciaries.

     

    2. Can employees use ChatGPT or Claude at work?

    Yes. There is no blanket ban, but organisations must control how personal data is processed. Staff should use approved tools and follow internal policy.

     

    3. Can HR upload CVs to an AI tool?

    Not to unapproved tools. First assess the purpose, lawful ground, vendor practices, safeguards and internal approvals.

     

    4. Is employee consent always required?

    No. Section 7(i) recognises legitimate uses for employment and safeguarding the employer. Where consent is relied on, it must meet the Act's standards.

     

    5. Is an AI vendor a Data Processor?

    Possibly, if it processes data on your behalf and instructions. Review the actual arrangement, not the vendor's label.

     

    6. Can we use overseas AI tools?

    There is no blanket prohibition. Section 16 allows restrictions on transfers to specified countries, and sector rules and contracts may also apply.

     

    7. What if an employee shares personal data with an AI tool by mistake?

    Start incident response with IT security and legal: assess the exposure, contain it, document it and decide on notifications.

     

    8. What are the penalties?

    The Schedule provides for up to ₹250 crore for failing to take reasonable security safeguards and up to ₹200 crore for failing to notify a breach, depending on the circumstances.

     

    9. Is a Data Protection Officer mandatory?

    Only for Significant Data Fiduciaries. Other organisations should still assign clear internal ownership.

     

    10. When do the main obligations apply?

    The main operational provisions commence on 13 May 2027. Some institutional provisions commenced earlier, so track each date.

     

    Conclusion

     

    The DPDP Act has made personal data protection part of everyday governance. As employees use generative AI for recruitment, communication and performance work, the goal is not to restrict AI but to set clear boundaries so people can benefit from it while employee, candidate and customer data stays protected. Begin preparing now for the main provisions due on 13 May 2027.

     

    Build AI capabilities alongside data protection awareness. Vinsys supports organisations through role-based AI training that connects emerging technologies with practical business workflows and responsible usage, helping teams adopt AI with confidence and accountability.

    DPDP Act employee AI usageDPDP Act for HRDPDP Act and employee dataAI usage policy for employeesemployee AI policy templateDPDP Act and generative AICan employees use ChatGPT at work under the DPDP Act?
    Individual and Corporate Training and Certification Provider
    VinsysLinkedIn16 November, 2026

    Vinsys Top IT Corporate Training Company for 2025 . Vinsys is a globally recognized provider of a wide array of professional services designed to meet the diverse needs of organizations across the globe. We specialize in Technical & Business Training, IT Development & Software Solutions, Foreign Language Services, Digital Learning, Resourcing & Recruitment, and Consulting. Our unwavering commitment to excellence is evident through our ISO 9001, 27001, and CMMIDEV/3 certifications, which validate our exceptional standards. With a successful track record spanning over two decades, we have effectively served more than 4,000 organizations across the globe.

    Table of Content
    DPDP timeline: what applies now and what comes nextDoes the DPDP Act apply to employee data?Why employee AI usage creates DPDP risk7 Steps Action Plan for HRTraining employees on AI and data protectionWho should own AI and DPDP Compliance
    Related Blogs
    India's DPDP Rule 2025: What It Means for Organizations and How to Get Compliance-Ready

    India's DPDP Rule 2025: What It Means for Organizations and How to Get Compliance-Ready

    DPDP Act Compliance Deadline: What Indian Enterprises Must Have in Place Before November 2026

    DPDP Act Compliance Deadline: What Indian Enterprises Must Have in Place Before November 2026

    Claude AI Corporate Training For Employees: Complete Guide for HR and L&D Teams

    Claude AI Corporate Training For Employees: Complete Guide for HR and L&D Teams

    ROI of Claude Training: How HR Can Justify the Budget

    ROI of Claude Training: How HR Can Justify the Budget

    Claude Corporate Training for Telecom, Retail, FMCG & E-commerce Teams

    Claude Corporate Training for Telecom, Retail, FMCG & E-commerce Teams

    Claude for MSMEs and Startups: Affordable AI Training for Small Teams

    Claude for MSMEs and Startups: Affordable AI Training for Small Teams

    Contact Us
    India
    United Arab Emirates
    United States of America
    Saudi Arabia
    Qatar
    Oman
    United Kingdom
    Republic Of The Congo
    Important Links
    • About Us
    • Investor
    • Career
    • CSR
    • Press Release
    • Contact Us
    Enquire
    • Emailenquiry@vinsys.com
    Stay Connected
    • Facebook
    • Twitter
    • LinkedIn
    • YouTube
    ©1998-2026 Vinsys | All Rights Reserved. Privacy Policy | Terms & Conditions
    X
    Select Language
    X