EU AI Act for Indian Companies: What Applies Now and What Was Deferred
EU AI Act for Indian Companies: What Applies Now and What Was Deferred
Consider a software company incorporated in Noida. It has no European office, no European employees, and no European servers. Its flagship product includes a conversational AI feature that answers customer queries on behalf of a distributor in Hamburg. Under Regulation (EU) 2024/1689, the European Union's Artificial Intelligence Act, that company is regulated by Brussels today. The statute attaches to where the output of an AI system is used, not to where its maker is incorporated.
Much of the commentary circulating in India still treats 2 August 2026 as the date on which the Act's high-risk obligations would begin to bite. That date has moved. Regulation (EU) 2026/1744, the Digital Omnibus on AI, entered into force on 27 July 2026 and deferred the core high-risk regime to 2 December 2027. Enterprises planning against the superseded calendar are allocating compliance budgets to the wrong years.
This article maps the law as it stands in November 2026: which provisions already bind Indian enterprises, which were deferred, and which obligations now travel through commercial contracts regardless of statutory timelines. It is structured around the questions Indian technology enterprises are actually asking their counsel.
In this analysis:
Section 1: Extraterritorial scope under Article 2
Section 2: The post-Omnibus compliance calendar and penalties
Section 3: Provider, deployer, and the Article 25 rebranding trap
Section 4: The Indian compliance stack: DPDP, GDPR, and domestic AI policy
Section 5: Sequencing compliance work before December 2027
Section 1: The Long Arm of Brussels: Extraterritorial Scope Under Article 2
The Output-Use Test
Article 2(1) of the AI Act defines the regulation's jurisdictional perimeter. It captures providers placing systems on the Union market, deployers established in the Union, and, critically for Indian enterprises, persons in third countries whose systems produce output used in the Union. The statutory formulation is unambiguous:
QUOTE: providers and deployers of AI systems that have their place of establishment or are located in a third country, where the output produced by the AI system is used in the Union.
The test requires no European subsidiary, no European data centre, and no European revenue threshold. A recruitment screening tool built in Bengaluru that ranks candidates for a Dutch employer, a credit analytics engine built in Pune whose scores inform a French lender, and a generative chatbot built in Gurugram that converses with Spanish consumers all satisfy it. Geography of incorporation is irrelevant; geography of effect is everything.
Why Indian Technology Enterprises Are Squarely in Scope
India is among the largest exporters of software services to the European Union, and its technology sector increasingly sells products rather than hours. Every Indian SaaS platform with EU customers, every global capability centre building AI features for a European parent, and every startup fine-tuning a foundation model for a European client sits within the perimeter Article 2 draws.
Enforcement against a non-EU entity is institutionally real rather than theoretical. The European AI Office now holds exclusive competence over AI systems built on general-purpose AI models by the same provider, with investigation and fining powers active since 2 August 2026. Member state market surveillance authorities, operational since August 2025, handle the remainder.
Section 2: The Compliance Calendar After the Digital Omnibus
What Already Applies: Prohibitions, Literacy, GPAI, Transparency
Four layers of the Act are live today. The prohibited practices of Article 5, covering manipulative techniques, social scoring, and specified biometric uses, have been unlawful since 2nd February 2025. The AI literacy duty of Article 4 applies to every provider and deployer, though the Digital Omnibus softened it from an obligation to ensure a given literacy level to an obligation to take measures supporting its development.
Obligations for general-purpose AI models, including technical documentation, copyright compliance policies, and public summaries of training data, have bound their providers since 2nd August 2025. Since 2nd August 2026, the transparency duties of Article 50 require disclosure when a person interacts with an AI system, labelling of synthetic content, and machine-readable marking of AI-generated output.
For an Indian enterprise, Article 50 is the provision most likely to bite first. Any chatbot, voice assistant, or content generation feature exposed to EU users must identify itself as artificial and mark its output. Systems placed on the market before 2nd August 2026 receive a short grace period: their marking obligations become enforceable on 2nd December 2026, the same date on which two new prohibitions, covering non-consensual intimate imagery and child sexual abuse material, take effect.
What Was Deferred: Annex III and the 2027 Deadline
The Digital Omnibus moved the high-risk regime, and only that regime. Obligations for systems listed in Annex III, which include AI used in recruitment, creditworthiness assessment, education admissions, and critical infrastructure, now apply from 2nd December 2027 rather than 2nd August 2026. High-risk systems embedded in regulated products under Annex I follow on 2nd August 2028. Nothing already in force was rolled back; the deferral is a fixed backstop, not an open-ended delay.
Compliance Calendar (consolidated, post-Omnibus):
Obligation | Provision | Application Date
Prohibited practices | Article 5 | 2nd February 2025
AI literacy measures | Article 4 | 2nd February 2025
General-purpose AI model duties | Articles 53-55 | 2nd August 2025
Transparency and labelling | Article 50 | 2nd August 2026
NCII and CSAM prohibitions | Article 5(1)(ba)-(bb) | 2nd December 2026
Legacy system output marking | Article 50(2) | 2nd December 2026
Annex III high-risk regime | Articles 6-49 | 2nd December 2027
Annex I product-embedded systems | Article 6(1) | 2nd August 2028
The enforcement architecture carries graduated penalties under Article 99. Violations of the Article 5 prohibitions attract fines of up to EUR 35 million or 7 percent of total worldwide annual turnover, whichever is higher. Breaches of most other obligations attract up to EUR 15 million or 3 percent, and supplying misleading information to authorities up to EUR 7.5 million or 1 percent. Worldwide turnover means exactly that: the Indian parent company's global revenue is the denominator.
Section 3: Provider, Deployer, or Both: Role Mapping for Indian Enterprises
The Statutory Roles
The Act allocates obligations by function, not by corporate label. A provider develops an AI system, or has it developed, and places it on the market under its own name or trademark. A deployer uses a system under its own authority in a professional context. An Indian SaaS company selling an AI-enabled product to EU customers is a provider. The same company, using a third-party recruitment tool internally, is a deployer. Most technology enterprises occupy both roles simultaneously, across different systems.
The Article 25 Rebranding Trap
Article 25 contains the provision most likely to surprise an Indian enterprise. A party that places its name or trademark on a high-risk system, makes a substantial modification to it, or redirects a general-purpose system into a high-risk use is deemed the provider, inheriting the full weight of provider obligations: risk management, data governance, technical documentation, conformity assessment, CE marking, and registration in the EU database.
An Indian company that white-labels a European or American AI engine for EU customers therefore does not escape the regime; it assumes its most demanding tier. The Digital Omnibus added a partial safeguard: the original provider must cooperate with the new provider and supply the technical information and access reasonably needed to complete conformity assessment. Whether that cooperation arrives in practice is a contractual question, not a statutory certainty.
Deployer Duties Under Article 26
Deployer obligations are lighter, but not trivial. Article 26 requires use in accordance with the provider's instructions, assignment of human oversight to persons with the competence, training, and authority to exercise it, retention of automatically generated logs for at least 6 months, and continuous monitoring for anomalies. Serious incidents must be escalated to the provider without delay, and use must be suspended where a risk to health, safety, or fundamental rights emerges.
Two further duties carry particular weight for employers. Deployers must inform workers before subjecting them to a high-risk system, and must inform individuals affected by its outputs. Deployers of credit-scoring or insurance-pricing systems, and public bodies generally, must additionally complete a fundamental rights impact assessment under Article 27 before first use.
Role Map:
Dimension | Provider | Deployer
Core duty | Build conformity into the system | Operate the system lawfully
Documentation | Technical file, conformity assessment, CE marking, EU database registration | Instructions followed, logs retained 6 months
Human oversight | Design the system to enable it | Assign it to trained, empowered personnel
Typical Indian entity | SaaS vendor, AI product startup, GCC building for the EU market | Enterprise using third-party AI tools internally
Section 4: The Indian Compliance Stack: DPDP, GDPR, and the Domestic AI Framework
The Contractual Transmission Mechanism
Statutory enforcement against a company with no European assets may be slow; contractual enforcement is immediate. European customers are already inserting AI Act compliance clauses into master services agreements, data processing agreements, and procurement terms, demanding evidence of conformity, Article 50 labelling commitments, and cooperation duties in the event of a regulatory inquiry.
For the Indian vendor, the operative question is therefore not whether Brussels can reach it directly, but whether its largest European customer will renew a contract without compliance documentation. This mirrors the dynamic the firm examined in the context of DPDP vendor liability: regulatory risk travels down the supply chain through paper before it travels through regulators.
India's Domestic Trajectory
The European regime does not operate in a vacuum. India's own framework is taking shape through the India AI Governance Guidelines released by MeitY on 5 November 2025, which adopt a voluntary-first, phased approach anchored in existing statutes, with the AI Safety Institute as the operational arm for testing and standards. The Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026, effective 20th February 2026, already impose labelling and due diligence duties for synthetically generated information.
The New Delhi Declaration on AI Impact, endorsed by 88 countries in February 2026, signals India's intent to shape global AI governance rather than merely receive it. For the compliance planner, the practical consequence is convergence: the vocabulary of Indian policy (transparency, provenance, accountability) increasingly mirrors the vocabulary of European law.
The compliance stack for an Indian technology enterprise is therefore tripartite. The DPDP Act, 2023 governs the personal data that trains and feeds AI systems; the GDPR governs the processing of EU personal data; and the AI Act overlays system-level obligations on top of both. Enterprises that treat these as separate projects will duplicate effort and miss the intersections, such as the new Article 4a legal basis for processing special-category data to detect and correct bias.
Section 5: Strategic Outlook: Compliance as Commercial Infrastructure
Sequencing the Work Before December 2027
The deferral to 2 December 2027 is best read as a construction window, not a reprieve. The rational sequence begins with an inventory of every AI system the enterprise provides or deploys, classified against the Act's taxonomy and screened against Article 2's output-use test. Role mapping follows: provider, deployer, or both. Transparency controls under Article 50 are already due and carry the earliest enforcement exposure. Contractual commitments to European customers must then be reconciled with statutory obligations, because commercial timelines will run ahead of regulatory ones.
The Broader Pattern of European Regulatory Reach
The AI Act continues a pattern Indian exporters already know from the GDPR and the Carbon Border Adjustment Mechanism: European regulation that projects its standards outward through market access rather than territorial jurisdiction. As the India-EU trade and investment framework deepens, that projection will intensify, and the enterprises best positioned within it will be those that treated compliance as infrastructure.
Enterprises that build compliance architecture once, and build it to the most demanding standard in their market footprint, convert regulation into a commercial asset. Those that wait for a domestic Indian mandate will find that their European customers have already imposed one. The question for Indian technology enterprises is no longer whether European AI law applies to them; it is whether their compliance posture will be ready when their customers ask for proof.
Strategic Legal Counsel
Discuss the implications of this briefing for your specific corporate or cross-border operations.