EU & cross-borderUpdated 22 November 20269 min read

EU AI Act High-Risk Rules for Indian HR-Tech and Credit Scoring

When Indian HR-tech and credit scoring products become high-risk AI under EU AI Act Annex III, what providers and deployers must do, the August 2026 date and the pending Digital Omnibus delay proposal.

By Suhas Bhairav · Reviewed against the DPDP Rules, 2025 (G.S.R. 846(E), 13 Nov 2025)

Legal basis
EU AI Act, Annex III
HR use cases
Recruitment, selection, promotion, termination, task allocation, monitoring
Credit use case
Creditworthiness and credit scoring of natural persons (fraud detection excepted)
Applies from
2 August 2026 (as adopted)
Pending change
Digital Omnibus proposal (Nov 2025) to delay some high-risk deadlines
Maximum fine
€15m or 3% of worldwide turnover

Quick answer

Under EU AI Act Annex III, AI used for recruitment, selection, promotion, termination, task allocation or monitoring of workers, and for creditworthiness assessment or credit scoring of individuals (except fraud detection), is high-risk. Indian providers selling such tools into the EU must meet risk management, data governance, documentation, logging, human oversight and conformity duties from 2 August 2026, unless the Digital Omnibus delay is adopted.

  • Annex III makes AI for recruitment, selection, promotion, termination, task allocation and worker monitoring high-risk.
  • AI for creditworthiness assessment or credit scoring of natural persons is high-risk, except systems used to detect financial fraud.
  • Indian providers of these tools to the EU need full high-risk compliance and an EU authorised representative.
  • Annex III obligations apply from 2 August 2026 as adopted; the November 2025 Digital Omnibus proposal would delay some high-risk deadlines if adopted.
  • In India, the same tools must meet DPDP duties from 13 May 2027, and banks face RBI FREE-AI expectations.

Who this is for: Indian HR-tech and recruitment platforms, Applicant tracking and assessment vendors, Credit scoring and lending-tech providers, Workforce analytics and monitoring tools, Product, ML and compliance leads

When is HR-tech or credit scoring high-risk under the AI Act?

Annex III of the EU AI Act lists use cases that make an AI system high-risk. Two matter most to Indian product companies.

Employment, workers' management and access to self-employment

AI systems intended to be used for:

  • recruitment or selection — placing targeted job ads, analysing and filtering applications, evaluating candidates;
  • decisions affecting terms of work relationships, promotion or termination;
  • allocating tasks based on individual behaviour or personal traits;
  • monitoring and evaluating performance and behaviour.

Access to essential private services

AI systems intended to evaluate the creditworthiness of natural persons or establish their credit score — with the exception of systems used to detect financial fraud.

Quick classification

Product featureLikely status
CV parser that ranks or filters candidatesHigh-risk
Video interview scoringHigh-risk (and check emotion recognition prohibitions)
Job-ad targeting by candidate profileHigh-risk
Shift scheduling based on worker behaviourHigh-risk
Productivity monitoring dashboardHigh-risk
Payroll calculationNot Annex III
Consumer credit scoring modelHigh-risk
Transaction fraud detectionExcluded from the credit category
Business (company) credit ratingNot "natural persons"; check carefully

The Act also contains a filter that can take a listed system out of high-risk status where it does not pose a significant risk — for example, performing a narrow procedural task — but systems that profile individuals remain high-risk. Document any reliance on it.

When do the obligations apply?

Under the Regulation as adopted, Annex III high-risk obligations apply from 2 August 2026. (High-risk AI embedded in Annex I regulated products follows on 2 August 2027.)

The Digital Omnibus

In November 2025, the European Commission proposed a Digital Omnibus that would delay some high-risk deadlines. A proposal is not law. It must be adopted by the European Parliament and the Council and published before any date changes.

Because today is after the original 2 August 2026 date, you need to know which regime you are in:

  1. Check the status of the Omnibus on EUR-Lex and the Commission's AI Act page.
  2. If it has not been adopted with a delay, assume Annex III obligations apply now.
  3. If it has, read the final text: delays may be conditional or limited to some obligations.

Already in force

Regardless of the Omnibus, prohibited practices have applied since 2 February 2025. Some emotion recognition uses in the workplace fall within them, which directly affects interview-analysis and employee-sentiment products. Check the prohibitions list before shipping such features.

What providers of high-risk systems must do

If your Indian company develops the HR or credit tool and places it on the EU market under its own name, you are the provider. The core requirements:

RequirementWhat to build
Risk management systemContinuous process to identify, evaluate and mitigate risks across the lifecycle
Data and data governanceTraining, validation and test data that are relevant, representative and examined for bias
Technical documentationFull description of system, design choices, data, testing and performance
Record-keepingAutomatic logging of events to enable traceability
Transparency to deployersInstructions for use: capabilities, limitations, accuracy, human oversight measures
Human oversightDesign enabling deployers to understand, monitor, override and stop the system
Accuracy, robustness, cybersecurityDeclared accuracy metrics; resilience to errors and attacks
Quality management systemDocumented processes for design, testing, change and post-market monitoring
Conformity assessmentAssessment before placing on the market, with declaration and CE marking
RegistrationRegistration in the EU database before placing on the market
Post-market monitoring and incident reportingOngoing monitoring; reporting serious incidents

Non-EU providers must also appoint an EU authorised representative.

Engineering translation

  • Bias testing across protected characteristics relevant in the EU, with documented thresholds.
  • Explanations a recruiter or credit officer can act on, not just SHAP plots.
  • Immutable decision logs: input features, model version, score, threshold, human action.
  • Override buttons that are real, logged and monitored.

What your EU customers (deployers) must do — and need from you

EU employers and lenders using your tool are deployers. Their duties shape your product requirements.

Deployers of high-risk systems must, among other things:

  • use the system according to the provider's instructions;
  • assign human oversight to competent, trained people;
  • ensure input data under their control is relevant and representative;
  • monitor operation and inform the provider of risks or serious incidents;
  • keep logs generated by the system for an appropriate period;
  • in the workplace, inform workers and their representatives before putting the system into use;
  • inform individuals that they are subject to a high-risk AI system when it makes or assists decisions about them;
  • for certain deployers, including those assessing creditworthiness, carry out a fundamental rights impact assessment.

Product features customers will ask for

  1. Admin controls for oversight roles and training records.
  2. Exportable logs with configurable retention.
  3. Candidate and applicant notices they can adapt.
  4. Monitoring dashboards showing drift and outcome disparities.
  5. Incident reporting channel to you.

GDPR also applies: recruitment and credit profiling usually require a DPIA under Article 35, and automated decision-making rules may apply. See AI hiring tools and the DPDP Act for the Indian side.

The same tools in India: DPDP and RBI

Many Indian HR-tech and lending-tech companies sell the same product at home. Indian law approaches it differently, but there is substantial overlap.

DPDP Act (from 13 May 2027)

  • Ground for processing. Employers can rely on the section 7 employment legitimate use for many HR purposes; recruitment of candidates who are not yet employees needs careful analysis, often consent.
  • Notice under Rule 3 for consent-based processing: itemised data and purposes, including AI assessment.
  • Security under Rule 6 and breach intimation under Rule 7.
  • Accuracy. Section 8(3) requires completeness, accuracy and consistency where data is used to make a decision affecting the Data Principal — directly relevant to scoring.
  • Algorithmic due diligence. If a customer is notified as a Significant Data Fiduciary, Rule 13(3) requires due diligence that algorithmic software is not likely to pose a risk to Data Principals' rights. See algorithmic due diligence.

RBI (credit scoring)

RBI's August 2025 FREE-AI report recommends risk-based classification, model validation, customer disclosure and human escalation for regulated entities. Binding effect depends on RBI directions; check current circulars.

One build for both markets

Bias testing, logging, human oversight and explanation features built for the AI Act also serve DPDP accuracy duties and FREE-AI expectations. Build once, configure per market.

Action plan and common mistakes

Action plan

  1. Classify each feature against Annex III; record reasoning, including any reliance on the significant-risk filter.
  2. Check prohibited practices for emotion recognition and similar features.
  3. Confirm the Omnibus status and your applicable date.
  4. Stand up a quality management system covering design, data, testing, change control and post-market monitoring.
  5. Produce technical documentation and instructions for use.
  6. Implement logging, human oversight and explanation features.
  7. Plan conformity assessment, EU database registration and CE marking.
  8. Appoint an EU authorised representative.
  9. Update customer contracts to allocate provider and deployer responsibilities.
  10. Align with GDPR DPIAs and DPDP duties for each market.

Common mistakes

  • Calling a candidate-ranking feature "decision support" to avoid classification. Assisting decisions about people is still in scope.
  • Treating fraud detection and credit scoring as the same; only fraud detection is excluded.
  • Forgetting worker monitoring and task allocation tools.
  • Waiting for the Omnibus without a fallback plan.
  • Building human oversight as a checkbox rather than a real ability to override.

For overall scoping, see EU AI Act for Indian companies.

Further reading on suhasbhairav.com

Before 13 May 2027

Compliance checklist

  1. Classify features against Annex III

    Map every HR and credit feature to Annex III categories and record the outcome.

    AI Act Annex III

  2. Screen for prohibited practices

    Check emotion recognition and similar features against the prohibitions in force since 2 Feb 2025.

    AI Act (prohibited practices)

  3. Confirm applicable date

    Check whether the Digital Omnibus has been adopted and what it changes.

    Digital Omnibus proposal

  4. Set up a quality management system

    Document design, data, testing, change and post-market processes.

    AI Act (high-risk providers)

  5. Test data for bias

    Assess representativeness and outcome disparities and document mitigations.

    AI Act (data governance)

  6. Build logging and oversight

    Log decisions immutably and give deployers real override and stop controls.

    AI Act (record-keeping, human oversight)

  7. Prepare instructions for use

    Describe capabilities, limitations, accuracy and oversight requirements for deployers.

    AI Act (transparency to deployers)

  8. Plan conformity and registration

    Schedule conformity assessment, CE marking and EU database registration.

    AI Act (high-risk providers)

  9. Appoint an EU authorised representative

    Required for non-EU providers of high-risk systems.

    AI Act

  10. Align with DPDP accuracy duty

    Ensure data used for scoring is complete, accurate and consistent for Indian users.

    DPDP s.8(3)

Key dates

  1. AI Act prohibited practices apply, including certain workplace emotion recognition

  2. Annex III high-risk obligations apply under the Regulation as adopted

  3. DPDP core obligations apply in India

  4. Annex I product-embedded high-risk obligations apply

Answers

Frequently asked questions

Is AI recruitment software high-risk under the EU AI Act?

Yes. Annex III lists AI intended for recruitment or selection, including targeted job ads, filtering applications and evaluating candidates, as high-risk. Indian providers selling such tools into the EU must meet high-risk requirements, including risk management, documentation, logging, human oversight and conformity assessment.

Is credit scoring high-risk under the AI Act?

AI used to evaluate creditworthiness or establish the credit score of natural persons is high-risk under Annex III, except systems used to detect financial fraud. Business credit ratings of companies are not about natural persons and need separate analysis.

When do high-risk rules for HR-tech apply?

Under the AI Act as adopted, Annex III obligations apply from 2 August 2026. The Commission's November 2025 Digital Omnibus proposal would delay some high-risk deadlines, but only if adopted. Check its status on EUR-Lex before assuming any delay.

Can an Indian HR-tech company avoid high-risk status?

Only if the system genuinely falls outside Annex III or meets the Act's filter for systems that do not pose a significant risk, such as narrow procedural tasks. Systems that profile individuals remain high-risk. Document the assessment; relabelling a ranking feature as decision support does not change its status.

Are emotion recognition interview tools allowed in the EU?

Emotion recognition in the workplace is among the AI Act's prohibited practices, with limited exceptions such as medical or safety reasons, and prohibitions have applied since 2 February 2025. Video interview tools that infer candidates' emotions may therefore be prohibited outright. Take legal advice before offering them.

What do EU employers need from an Indian HR-tech vendor?

Instructions for use, human oversight controls, exportable logs, notices for workers and candidates, monitoring dashboards and an incident channel. Deployers must inform workers' representatives before workplace use and tell individuals when a high-risk system assists decisions about them.

Legal references and sources

  • EU AI Act, Regulation (EU) 2024/1689 — Annex III (employment; creditworthiness) (official text)
  • EU AI Act, Regulation (EU) 2024/1689 — High-risk requirements and penalties (official text)
  • GDPR, Regulation (EU) 2016/679 — Article 35 (DPIA) (official text)
  • Digital Personal Data Protection Rules, 2025 — Rules 6, 7 and 13(3) (official text)

This guide explains the law for engineering and business teams. It is not legal advice; confirm how it applies to your organisation with qualified counsel, and check the Gazette for later notifications or amendments.

About the author

Suhas Bhairav

Suhas Bhairav

Systems architect and applied AI practitioner who designs production AI systems for Indian and global companies, with a focus on privacy-by-design under the DPDP Act and the EU AI Act.