Tax and Corporate Compliance for AI Businesses in India: An Astro247 Case Study

#AD

Artificial intelligence is changing how businesses are built and services are delivered in India, but an AI-driven company does not operate under a separate tax or corporate-law framework merely because it uses AI. Its obligations continue to arise from existing laws governing companies, income tax, GST, cross-border payments, digital transactions, personal data and consumer protection.

What makes compliance more difficult is that several of these frameworks can apply to the same business activity. A consumer-facing AI platform may be incorporated in India, rely on overseas cloud infrastructure, pay foreign AI-model providers, collect personal information from users, sell subscriptions in India and abroad, and automatically generate personalised content. Each of those activities can create a different compliance question.

The regulatory environment has also changed substantially. Angel tax has been withdrawn, the equalisation levy has been dismantled, the Income-tax Act, 2025, has replaced the Income-tax Act, 1961, from 1 April 2026, the Digital Personal Data Protection Rules were notified in November 2025, and India’s treatment of synthetically generated information developed further through amendments to the Information Technology Rules in February 2026.

The practical question for founders is therefore not, “What is the tax on AI?” It is how existing tax, corporate and digital laws apply to an AI-led operating model.

This article examines those considerations using an AI-based astrotech platform such as Astro247 as an illustrative example. The discussion reflects the legal and regulatory position reviewed as of September 2026 and is intended as general information rather than advice on any individual transaction.

1. Choosing the Right Business Structure

The legal structure of an AI business affects fundraising, governance, ownership and taxation. A private limited company is commonly used where founders expect to raise external capital, bring investors onto the cap table or establish an employee stock-option programme. An LLP can suit bootstrapped or founder-led businesses where ownership is comparatively simple, while a One Person Company may be relevant where a single entrepreneur is starting the venture.

The decision should be based on the actual business plan rather than assumptions about which structure is automatically “best” for a technology company. Important factors include:

• expected fundraising;
• number and type of owners;
• ESOP plans;
• governance obligations;
• tax treatment;
• future restructuring; and
• plans for international expansion.

Older guidance around OPCs also needs care. The Companies (Incorporation) Second Amendment Rules, 2021, removed the earlier mandatory conversion framework linked to specified capital and turnover thresholds and liberalised the conversion of OPCs into other company forms.

Once the entity is incorporated, technology does not reduce its normal corporate responsibilities. Depending on the structure and circumstances, these can include statutory records, board and shareholder processes, annual financial statements, annual returns and filings relating to directors, capital and other corporate changes. For an AI startup, the corporate structure should therefore be chosen with future business requirements in mind rather than merely the lowest incorporation cost.

2. Startup Recognition and the End of Angel Tax

India’s startup tax environment changed materially with the withdrawal of the former angel-tax provision. Section 56(2)(viib) of the Income-tax Act, 1961, commonly associated with angel tax, was sunset through the Finance (No. 2) Act, 2024, and does not apply from Assessment Year 2025-26.

However, its withdrawal does not affect proceedings or record-retention requirements relating to earlier assessment years. Startups that raised capital while the provision was in force should continue preserving valuation reports, share-subscription documents, investor details, board approvals and related records, particularly if an assessment, appeal or other proceeding remains open.

DPIIT recognition under Startup India may continue to be relevant even though angel tax has been removed. For periods governed by the Income-tax Act, 1961, the profit-linked deduction was available under Section 80-IAC. From Tax Year 2026-27, the corresponding provision is Section 140 of the Income-tax Act, 2025.

Subject to the prescribed conditions and approval requirements, eligible startups incorporated on or after 1 April 2016 but before 1 April 2030 may claim a deduction of 100 per cent of eligible business profits for three consecutive years out of the prescribed ten-year period.

Founders should therefore treat Startup India recognition as a compliance and tax-planning question rather than a branding exercise. The appropriate analysis is whether the company satisfies the current conditions and whether the available benefits fit its expected profitability, ownership and tax position.

3. Income-tax Compliance After the Income-tax Act, 2025

The direct-tax framework entered a new stage on 1 April 2026. The Income-tax Act, 2025, now applies from Tax Year 2026-27 and replaces the Income-tax Act, 1961. For businesses, this transition is more than a change of statute name. Accounting systems, tax templates, compliance calendars, contracts and internal documents that historically referred to provisions of the 1961 Act must now be read against the corresponding framework under the new law.

AI-driven companies should review areas such as:

• advance-tax calculations;
• withholding procedures;
• foreign remittances;
• return filing;
• tax software;
• internal finance policies; and
• contractual tax references.

The Income Tax Department has also introduced forms and challans under the new Act. Continuing to use old section references mechanically can therefore create avoidable compliance errors even where the commercial transaction itself has not changed.

For AI businesses, this transition is especially relevant because expenditure is often spread across domestic and international suppliers. A startup may simultaneously pay Indian employees, independent contractors, overseas SaaS companies, cloud providers, advertising platforms and AI-model vendors.

The company remains the taxpayer. The use of artificial intelligence does not, by itself, create a separate category of taxable business income. The practical focus remains on recording revenue correctly, supporting deductions, reviewing withholding obligations and filing under the tax framework applicable to the relevant tax year.

4. Payments to Foreign AI, Cloud and Software Providers

Cross-border technology payments are one of the areas where AI startups can create compliance risk without realising it. A foreign SaaS or AI invoice may look like an ordinary online subscription, but an Indian company should consider whether the payment creates withholding or remittance obligations. It would be incorrect to assume that every payment to an overseas technology company automatically attracts tax deduction at source.

The analysis depends on factors such as:

• nature of the service;
• contractual terms;
• country of the recipient;
• whether the payment is chargeable to tax in India;
• relevant domestic tax provisions; and
• any applicable Double Taxation Avoidance Agreement.

Payments for cloud infrastructure, API access, software subscriptions, technical services, analytics or advertising may therefore require different treatment. The foreign-remittance framework links withholding to sums chargeable to tax in India and also recognises the relevance of treaty provisions. A business should therefore avoid applying one standard withholding assumption to every overseas technology vendor.

From Tax Year 2026-27, Forms 145 and 146 operate as the new-Act equivalents of the foreign-remittance forms previously numbered Forms 15CA and 15CB. Form 145 contains the prescribed remittance information, while Form 146 is the accountant’s certificate required in applicable cases.

A practical internal control is to maintain a vendor-wise review recording the service, jurisdiction, contractual terms, tax analysis and supporting documentation. That is particularly useful where dozens of recurring SaaS and AI subscriptions are paid automatically each month.

5. GST Treatment of AI-Driven Digital Services

GST treatment depends on what the customer is actually buying, not merely on the fact that artificial intelligence is involved. Many IT and digital services fall within the 18 per cent GST framework, and the broader GST rate rationalisation implemented from 22 September 2025 retained 18 per cent as the principal standard rate for services. An AI business should still determine the classification and applicable rate of its particular service rather than assuming that every digital product receives identical treatment.

Relevant questions include:

• What is the principal supply?
• How is the service delivered?
• To what extent is its supply dependent on information technology?
• Is there substantive human involvement?
• Where is the supplier located?
• Where is the customer located?
• Is the customer GST-registered?
• Is the transaction domestic, imported or exported?

These questions become important for subscription platforms because automated software delivery may interact with the rules governing Online Information and Database Access or Retrieval services. A consumer AI platform that sells automatically generated reports, recommendations or other digital output may therefore require a different GST analysis from a platform that primarily provides a professional service through a human adviser.

The correct approach is to analyse each product line by the way it is delivered. A business offering both automated and human-assisted services should not assume that one GST classification necessarily applies to the entire platform.

6. OIDAR and Automated AI Subscriptions

OIDAR, or Online Information and Database Access or Retrieval services, is particularly relevant to internet-delivered digital businesses. The GST framework broadly addresses services delivered over the internet or an electronic network where information technology is essential to the supply.

The statutory definition was amended in 2023 by removing the earlier reference to services being “essentially automated and involving minimal human intervention.” Accordingly, substantive human involvement does not, by itself, conclusively take a service outside OIDAR.

For an AI-driven subscription product, the actual delivery model remains critical. Suppose a customer enters information into an application and receives a personalised report generated automatically by software. That type of transaction raises a strong OIDAR question because the service is digitally supplied and its delivery is dependent on an automated information-technology system.

Human involvement may affect the factual analysis, but it does not by itself exclude a service from OIDAR after the 2023 amendment. The decisive question is whether the service is delivered over the internet or an electronic network and whether its supply is essentially dependent on information technology and cannot be ensured in the absence of information technology.

For example, if an astrotech platform provides one product where an AI system automatically generates a report and another where a human astrologer personally reviews the user’s information and delivers a consultation, the two offerings require separate analysis based on their actual characteristics. The mere presence of a human adviser does not automatically determine the classification.

It is therefore safer to say that an AI subscription may qualify as OIDAR depending on its characteristics rather than claiming that all AI subscriptions are automatically OIDAR. Product design, contractual description and actual delivery should all be considered before deciding the correct GST treatment.

7. Imported Services and Reverse-Charge Considerations

AI businesses frequently purchase services from companies located outside India. These may include cloud hosting, AI-model access, APIs, analytics software, payment technology, marketing tools and other SaaS products.

Where a person located in India receives a taxable supply of services from a supplier located outside India and the transaction satisfies the statutory conditions for an import of services, IGST is generally payable by the recipient under reverse charge, subject to applicable exemptions and special provisions.

The fact that the service is purchased online or charged automatically to a corporate card does not remove the need to determine whether it constitutes an import of services for GST purposes. The implications should also not be examined solely by reference to whether the recipient is already registered, as liability under reverse charge may have separate registration consequences in applicable cases.

The finance team should therefore reconcile foreign-vendor payments with the GST compliance process rather than reviewing imported services only at year-end. Where input tax credit is otherwise available, the business may claim eligible credit subject to the statutory conditions and restrictions. However, the underlying reverse-charge liability must first be identified, discharged, recorded and reported correctly.

A practical monthly process can include:

• reviewing overseas vendor invoices;
• matching them to bank or card payments;
• identifying the nature of the service;
• determining whether the transaction constitutes an import of services;
• checking whether reverse charge applies;
• recording and paying the applicable IGST; and
• reviewing input-tax-credit eligibility.

For an AI company with a large software stack, this procedure can prevent dozens of small recurring foreign subscriptions from becoming a larger compliance problem later.

8. Overseas Customers and Export-of-Service Rules

An Indian AI business serving users outside India should not assume that every foreign-customer transaction is automatically outside GST. To receive zero-rated treatment, the transaction must satisfy the statutory conditions for export of services.

These conditions generally require that:

• the supplier of the service is located in India;
• the recipient is located outside India;
• the place of supply is outside India;
• payment is received in convertible foreign exchange or in Indian rupees wherever permitted by the Reserve Bank of India; and
• the supplier and recipient are not merely establishments of the same person.

Where the applicable conditions are satisfied, the export of services can be treated as a zero-rated supply. If exports are made without payment of IGST, the business must also follow the applicable Letter of Undertaking procedure.

For a digital subscription platform, documentation becomes especially important because the customer may never interact physically with the supplier. The business should retain reliable evidence relating to:

• customer location;
• billing information;
• payment receipt;
• invoices;
• contractual relationship; and
• nature of the service supplied.

These records support both GST classification and the company’s wider tax position.

A payment gateway may make an international transaction appear simple from the customer’s perspective, but the tax analysis behind that payment still depends on statutory conditions. International expansion should therefore include a tax review rather than being treated only as a marketing or payments decision.

9. Equalisation Levy After Its Withdrawal

The equalisation levy is now primarily a historical compliance issue rather than a current charge for most AI businesses. The 2 per cent equalisation levy applicable to specified e-commerce supplies was withdrawn with effect from 1 August 2024. The earlier 6 per cent levy on specified services, including certain online advertising arrangements, was subsequently removed from 1 April 2025.

This matters because older compliance guides may still instruct businesses to consider the equalisation levy when paying certain overseas digital platforms. Current transactions should not be processed automatically under a levy that has been withdrawn.

At the same time, businesses should not treat withdrawal as permission to discard older records. Historical liabilities, filings and supporting documents relating to periods during which the levy was legally applicable may still be relevant.

AI startups with significant overseas advertising or technology spending should therefore review their finance procedures to ensure that legacy equalisation-levy assumptions have been removed from current workflows while historical documentation remains preserved.

10. DPDP Compliance and the Role of the Data Fiduciary

Personal-data compliance is becoming a core operational issue for consumer-facing AI businesses. The Digital Personal Data Protection Act, 2023, established India’s principal framework for digital personal-data processing, while the Digital Personal Data Protection Rules, 2025, were notified in November 2025 with phased commencement.

This means businesses should avoid two extremes. It would be wrong to claim that every substantive obligation became fully enforceable immediately in November 2025, but it would be equally risky to ignore the framework until later commencement dates arrive.

Every obligation should therefore be matched with its notified commencement date. Provisions scheduled to commence later should be treated as implementation-readiness requirements until they become legally operative.

Product changes take time. Consent mechanisms, notices, vendor agreements, security procedures and retention workflows may all require technical implementation. A business that determines why and how personal data is processed can fall within the concept of a Data Fiduciary. Vendors processing data on its behalf may function as Data Processors, depending on the contractual and factual arrangement.

The practical starting point is data mapping. The business should know:

• what personal data it collects;
• why it collects each category;
• where the data is stored;
• which third parties receive it;
• how long it is retained; and
• how users can exercise applicable rights.

The final DPDP Rules contain requirements for notices to identify the relevant personal data and specified purpose clearly. These requirements must be implemented in accordance with their notified commencement. A generic privacy policy cannot substitute for understanding the actual data flows of the product.

11. Children’s Data and Retention Require Separate Analysis

Children’s personal data deserves additional attention for consumer-facing AI platforms because a service may be accessible to users under 18 even where children are not the principal target audience.

The DPDP framework contains additional requirements relating to children’s data, including provisions addressing verifiable parental consent, subject to the applicable commencement dates, statutory conditions and exemptions. A platform should therefore examine whether its onboarding process can identify situations in which children’s-data obligations arise rather than relying only on a statement in its terms that minors should not use the service.

Data retention also requires precision. It would be inaccurate to state that every AI business must delete all inactive-user data after the same fixed period. The final DPDP Rules contain specified inactivity periods for certain defined categories of large digital platforms, while other organisations must consider the general statutory rules governing purpose, retention and erasure.

For an astrotech platform, information such as a user’s name, date of birth, contact details, account data and submitted questions may constitute personal data where it relates to an identifiable individual. The business should establish a documented reason for collecting each category, determine how long it is genuinely required and consider whether different data types need different retention periods.

12. AI Governance and Synthetically Generated Information

India’s AI-governance framework has continued to develop. MeitY and the IndiaAI Mission released the India AI Governance Guidelines in November 2025, reflecting an approach that relies on governance principles and existing laws rather than imposing one single statute on every AI system.

The regulatory position moved further in February 2026 when amendments to the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021, concerning synthetically generated information were notified on 10 February 2026 and came into force on 20 February 2026.

Importantly, the notified definition of synthetically generated information is confined to artificially or algorithmically generated or modified audio, visual or audio-visual information. Purely textual AI output does not fall within that definition merely because it is AI-generated.

The amendments contain obligations relevant to intermediaries and significant social media intermediaries, including provisions concerning declarations, technical measures and labelling in specified circumstances.

An important compliance distinction is that the existence of these rules does not mean every AI-generated output from every application automatically carries the same legal labelling requirement.

A company should first determine:

• whether it falls within the relevant category of intermediary;
• whether the content is audio, visual or audio-visual information covered by the definition;
• whether the particular information satisfies the statutory definition;
• what obligations apply to the company’s role; and
• whether additional transparency measures are appropriate even where a specific statutory label is not mandatory.

This prevents businesses from either overstating the law or ignoring genuinely applicable obligations. For AI companies, legal classification should come before designing compliance labels or product warnings.

13. Consumer Protection and AI-Generated Claims

AI businesses also need to consider how their products are represented to customers. The Consumer Protection Act, 2019, and the wider consumer-protection framework apply independently of whether a service uses artificial intelligence. Misleading representations or unsubstantiated claims can therefore create risk even where no AI-specific rule is involved.

Particular care may be required around claims concerning:

• accuracy;
• guaranteed outcomes;
• predictions;
• expertise;
• financial consequences;
• health-related effects; and
• the extent of human involvement.

This issue becomes more important for products where users may place significant reliance on automatically generated output. An AI astrology platform, for example, should distinguish between providing interpretive guidance and representing an AI-generated prediction as a guaranteed future event. Similar considerations can arise in financial, educational, legal-information and health-related AI products.

The wording surrounding the output is therefore part of compliance. A technically accurate algorithm does not protect a business if its marketing material overstates what the system can reliably deliver. AI companies should review landing pages, advertisements, onboarding screens and in-product messages together because the overall consumer impression can matter more than one isolated disclaimer.

14. The Astro247 Case Study

An AI-based Vedic astrology platform such as Astro247 provides a useful example because corporate, tax, GST, data and AI-governance questions can arise within one customer journey.

At the corporate and direct-tax level, the first question is the entity through which the business operates and whether any Startup India benefits are relevant. Revenue from subscriptions or paid services must be recorded under the current tax framework, while payments to overseas AI models, cloud providers or software vendors should be reviewed separately for withholding and remittance obligations.

GST analysis starts with the product. A fully automated report generated from user-submitted information may raise an OIDAR issue. A service involving substantive interaction with a human astrologer must also be independently examined under the amended OIDAR definition and other applicable classification provisions; human involvement alone does not settle the issue.

International customers may create export-of-services questions, while foreign technology inputs can create reverse-charge considerations.

The data layer is equally important. An astrotech platform may process names, dates of birth, contact information, account records and user-submitted questions. The business needs to understand why each category is collected, which systems and third parties receive it, how it is protected and how long it should be retained.

Finally, AI-generated outputs must be presented accurately. The platform should avoid unsupported claims of certainty and explain the nature of its service in a way that does not mislead users.

This discussion does not certify Astro247’s present compliance position or endorse the accuracy of its services or predictions. It illustrates how the legal framework may apply to an AI-driven astrotech business.

15. Building a Practical Compliance Framework

An early-stage AI company may not need a separate team for every area of regulation, but it does need clarity about who owns each compliance responsibility. Tax advisers cannot design data-retention architecture by themselves, and software engineers should not be expected to decide whether a payment to a foreign vendor is taxable in India.

A practical compliance map can help different functions work from the same business information.

Area	Key matters to review
Corporate	Entity structure, MCA filings, governance and statutory records
Income tax	Current tax framework, Section 140 startup benefit, advance tax and returns
Overseas payments	Chargeability, DTAA analysis, withholding and remittance records
GST	Classification, rate, invoicing, OIDAR and place-of-supply analysis
Imported services	Import-of-service conditions, reverse-charge liability and eligible input tax credit
Overseas customers	Export conditions, payment realisation and LUT requirements
Data protection	Commencement dates, data mapping, notices, consent, processors and retention
Children’s data	Age and parental-consent requirements where applicable
AI content	Intermediary status, SGI scope, IT Rules and transparency obligations
Consumer protection	Accuracy of claims and avoidance of misleading representations

This map should be revisited whenever the business launches a new product, enters another country, changes its payment model, uses a new AI provider or starts collecting a new category of personal data.

Conclusion

AI-driven businesses in India do not face a separate AI tax regime, but they operate across several overlapping legal frameworks. Company law, income tax, GST, cross-border payments, data protection, the Information Technology Rules and consumer protection can all affect the same business model.

For platforms such as Astro247, the practical approach is to map each compliance requirement to the actual product, customer journey, data flow and vendor relationship.

Recent changes, including the Income-tax Act, 2025, the phased implementation of the DPDP Rules, the withdrawal of the equalisation levy and the 2026 Information Technology Rules amendments, make older compliance templates less reliable.

Businesses that review these obligations early, with qualified tax, legal and data advisers, are better placed to reduce compliance risk, maintain reliable records and support sustainable growth.

*****

Disclaimer: This article is intended solely for general informational and educational purposes and does not constitute legal, tax, accounting, investment, data-protection or other professional advice. References to Astro247 are illustrative and do not certify its regulatory compliance or constitute an endorsement by TaxGuru of the platform, its services, predictions, claims or outcomes. Laws, rules, notifications and their commencement dates may change, and their application depends on the facts of each case. Readers should obtain advice from qualified professionals before acting on this information.