Short answer: yes, the EU AI Act can apply to an Indian SaaS company, and for most vendors the first job is to work out whether each AI feature makes you a provider or a deployer. The Act reaches providers that place AI systems on the EU market wherever they are established. It also reaches providers and deployers outside the EU where the system's output is used in the EU. An Indian vendor selling an AI feature to EU customers is usually the provider, and the EU customer is the deployer. The AI Omnibus moved the deadlines for high-risk systems, to 2 December 2027 for Annex III uses and 2 August 2028 for AI built into regulated products. The obligations did not go away. Transparency rules (Article 50) have applied since 2 August 2026.
This post is for the engineering and product owners at an India-based SaaS company. It covers who is who, which dates now apply, what to build and what to ask counsel. It is engineering guidance, not legal advice. Mono is not a law firm and does not certify anyone. Your EU counsel decides how the Act applies to your product.
Does the EU AI Act apply to an Indian SaaS company?
It can. Article 2(1) sets the reach. Point (a) covers providers placing AI systems on the market in the EU "irrespective of whether those providers are established" in a third country. Point (c) covers providers and deployers established in a third country "where the output produced by the AI system is used in the Union".
In practice, three facts decide it:
- Who uses the feature. If EU businesses or EU residents use your AI feature, you are placing a system on the EU market or your output is used there.
- Whether it is an AI system. The Act defines an AI system as a machine-based system that infers, from the input it receives, how to generate outputs such as predictions, content, recommendations or decisions. A rules-only report builder is probably not one. A summariser, a classifier or a scoring model probably is.
- What the feature does. A support chatbot, a hiring score and a document summariser sit in very different tiers. The tier decides how heavy your duties are.
Having no office in the EU does not take you out of scope. If the feature is high-risk, a vendor with no EU establishment must also appoint an authorised representative in the EU (covered below).
Are you a provider or a deployer?
Article 3 defines both roles. A provider develops an AI system or has one developed, and places it on the market or puts it into service "under its own name or trademark, whether for payment or free of charge". A deployer uses an AI system "under its authority", except in personal non-professional activity. The same company can hold both roles for different systems.
| Situation | Likely role | Engineering implication |
|---|---|---|
| You build a screening feature into your SaaS and sell it under your brand to EU customers | Likely role: Provider (you). The EU customer is the deployer | Engineering implication: You own the technical file, logs, instructions for use and human oversight design |
| You call a third-party model API inside your product | Likely role: Provider of your system. The model vendor is a separate actor upstream | Engineering implication: You still answer for how your system behaves. Keep the vendor's terms and model documentation in your evidence |
| Your own staff use a chatbot or coding assistant internally | Likely role: Deployer | Engineering implication: Use it as the provider's instructions say, train staff, and review outputs |
| You white-label another vendor's AI system under your brand | Likely role: May become the provider under Article 25 if the system is high-risk | Engineering implication: Contract for access to documentation, and ask counsel before you rebrand |
| You change a general-purpose tool so that it is used for a high-risk purpose | Likely role: May become the provider under Article 25(1)(c) | Engineering implication: Check the intended purpose in your product copy and sales material |
Article 25 matters to SaaS teams. A distributor, importer, deployer or other third party is treated as the provider of a high-risk AI system if it puts its name or trademark on one, makes a substantial modification to one, or changes the intended purpose of a non-high-risk system so that it becomes high-risk. Intended purpose is read from your instructions for use, your marketing and your sales materials, not just your code. So a product page that promises "automated candidate ranking" can shape your tier.
Which risk tier is your AI feature in?

The Commission describes four tiers: unacceptable risk (nine prohibited practices), high risk, transparency risk and minimal risk. Most SaaS features land in the last two. A few land in the high-risk tier, and that is where the work is.
| Tier | What it covers | What a SaaS vendor does |
|---|---|---|
| Prohibited | What it covers: Nine practices, such as the non-consensual sexual content practice that applies from 2 December 2026 under the Omnibus | What a SaaS vendor does: Do not build them |
| High-risk | What it covers: Annex III uses, including biometrics, education, employment and worker management, essential services such as credit scoring, and migration; and AI that is a safety component of regulated products (Annex I) | What a SaaS vendor does: Full provider duties: risk management, data governance, technical documentation, logs, human oversight, accuracy and cybersecurity, a quality management system, conformity assessment and registration |
| Transparency (Article 50) | What it covers: AI that talks to people, AI that generates synthetic audio, image, video or text, emotion recognition, and deepfakes | What a SaaS vendor does: Tell people they are dealing with AI, mark synthetic output in a machine-readable way, and give deployers what they need to disclose |
| Minimal | What it covers: Everything else, such as a spam filter or search ranking with no Annex III use | What a SaaS vendor does: No Act-specific duties beyond AI literacy, but keep the inventory |
Annex III lists the high-risk uses. Those most relevant to SaaS are recruitment and selection tools, systems that decide terms of work, promotion or termination, systems that monitor and evaluate worker performance, systems for admission to education and for evaluating learning outcomes, and credit scoring. A provider that thinks its Annex III system is not high-risk has to document that assessment before it places the system on the market (Article 6(4)) and register it (Article 49(2)). A system that profiles natural persons stays high-risk.
What did the AI Omnibus change, and what dates apply now?
The AI Omnibus (Regulation (EU) 2026/1744, OJ 24 July 2026) amended the Act. The Commission proposed it on 19 November 2025. Political agreement came on 7 May 2026, Parliament approved it on 16 June 2026 and the Council adopted it on 29 June 2026. It entered into force on 27 July 2026. The Act itself (Regulation (EU) 2024/1689) entered into force on 1 August 2024.
The Omnibus moved deadlines. Obligations remain. This table shows the position as of 20 November 2026 on the Commission's AI Act page and the Future of Privacy Forum's timeline.
| Date | What applies |
|---|---|
| 2 February 2025 | What applies: Prohibitions 1 to 8 and AI literacy |
| 2 August 2025 | What applies: Obligations for general-purpose AI model providers |
| 2 August 2026 | What applies: The Act became generally applicable, with exceptions. Transparency rules (Article 50) apply, and the AI Office's enforcement powers over general-purpose AI begin, including fines |
| 2 December 2026 | What applies: Prohibition 9 (non-consensual sexual content and child sexual abuse material, including "nudification") applies. Article 50(2) marking applies to certain existing systems |
| 2 August 2027 | What applies: General-purpose AI models placed on the market before 2 August 2025 must comply. Each Member State must have at least one regulatory sandbox |
| 2 December 2027 | What applies: Annex III high-risk systems (biometrics, education, employment, migration and so on). The original date was 2 August 2026 |
| 2 August 2028 | What applies: Annex I high-risk systems (AI in regulated products). The original date was 2 August 2027 |
Do not write "the high-risk rules apply from August 2026" in any customer document. That was the old date. Equally, do not tell a buyer you have until December 2027 to start. A high-risk system needs a risk management system, data governance, a technical file and logs, and those take months to build and test. Teams that start in 2027 will be rushing.
The Omnibus also softened the AI literacy duty. Article 4 now asks providers and deployers to take measures to support the development of AI literacy among staff and others who operate AI on their behalf, rather than to ensure it. It also extends simplified requirements to small mid-cap companies. We have not confirmed which provisions that covers, so ask counsel whether you qualify.
What must you do now, before the high-risk dates?
Four things apply or will soon apply to many SaaS vendors whatever the high-risk timetable:
- Transparency (Article 50). If your product interacts directly with people, they must be informed they are dealing with an AI system, unless that is obvious to a reasonably well-informed person. Providers of systems that generate synthetic audio, image, video or text must mark outputs in a machine-readable format and make them detectable as artificially generated. Deployers of systems that generate deepfake image, audio or video content must disclose that it is artificial. Build the disclosure into the UI and the output metadata, and give EU deployers the settings they need to meet their own duty.
- AI literacy (Article 4). Keep a record of the training and guidance you give to staff who build, run or sell AI features. The duty is lighter after the Omnibus, but a buyer's questionnaire will still ask.
- Prohibited practices (Article 5). Check your roadmap against the list. Nine practices are banned, and the ninth applies from 2 December 2026.
- General-purpose AI models. If you only call another company's model, the general-purpose model obligations generally sit with that company as the model provider. You still carry the duties for your own system. If you train or substantially retrain your own general-purpose model, take advice, because the model obligations may fall on you.
What does a high-risk provider have to build?

If one of your features falls in Annex III, treat the provider duties as a product backlog. The table maps the main Articles to things an engineering team can ship or document.
| Article | Requirement | What to build or keep |
|---|---|---|
| 9 | Requirement: Risk management system | What to build or keep: A risk register per feature, with tests, thresholds and a review cycle over the life of the system |
| 10 | Requirement: Data and data governance | What to build or keep: Documented training, validation and test data sources, quality checks and bias examination |
| 11 | Requirement: Technical documentation | What to build or keep: A technical file kept current: design, data, evaluation results, limits |
| 12 and 19 | Requirement: Record-keeping and logs | What to build or keep: Automatic event logs, retained and protected, that let you trace the system's behaviour |
| 13 | Requirement: Information for deployers | What to build or keep: Instructions for use: intended purpose, accuracy, known limits, human oversight measures |
| 14 | Requirement: Human oversight | What to build or keep: UI and workflow controls that let a trained person understand outputs, override them and stop the system |
| 15 | Requirement: Accuracy, robustness and cybersecurity | What to build or keep: Evaluation suites, adversarial and failure tests, access control and monitoring |
| 17 | Requirement: Quality management system | What to build or keep: Written procedures for design, testing, change control, data management, incidents and records |
| Articles 43, 47 to 49 | Requirement: Conformity assessment, declaration, marking and registration | What to build or keep: A signed declaration, the conformity marking and an entry in the EU database before market placement |
| 22 | Requirement: Authorised representative | What to build or keep: For a provider outside the EU: a written mandate with an EU-based representative before the system is made available |
| 72 and 73 | Requirement: Post-market monitoring and serious incidents | What to build or keep: A monitoring plan and an incident process that can meet the Article 73 clocks |
Much of this overlaps with what you build for an enterprise buyer anyway: tenant AI settings, audit logs, human review steps and a versioned technical file. That is why we treat the Act as an engineering backlog, not a paper exercise. A management system standard such as ISO 42001 or the NIST AI RMF can organise the work. Neither is a substitute for the Act's own requirements. Do not assume a certificate gives a presumption of conformity. Ask counsel which standards, if any, do.
How do serious incident reports work for a provider?
Providers of high-risk systems must report serious incidents to the market surveillance authority under Article 73. The clocks are 15 days from becoming aware as the general rule, 2 days for a widespread infringement or a disruption of critical infrastructure, and 10 days for a death. Article 3(49) defines a serious incident: death or serious harm to health, disruption of critical infrastructure, infringement of fundamental-rights obligations, or serious harm to property or the environment.
India has no AI-specific incident reporting rule as of November 2026, so a single incident can start several clocks at once, for example CERT-In (6 hours) and the DPDP Board report (72 hours from 13 May 2027) alongside Article 73. Build one incident record that feeds all of them. Our post on AI incident response covers the runbook.
What are the fines?
Under the AI Act as adopted in 2024, Article 99 sets three bands. The higher of the figure and the percentage applies, except that for SMEs the lower applies.
| Breach | Maximum fine |
|---|---|
| Prohibited practices | Maximum fine: €35 million or 7% of worldwide annual turnover |
| Most other obligations | Maximum fine: €15 million or 3% |
| Supplying incorrect or misleading information to authorities | Maximum fine: €7.5 million or 1% |
The Omnibus may have changed parts of Article 99. We have not verified that, so check the current text with counsel before you quote a figure to a customer or a board.
A 90-day readiness plan for an Indian SaaS vendor
This plan assumes you sell to EU customers or expect to. Adjust it with counsel.
| Weeks | Work | Output |
|---|---|---|
| 1 to 2 | Work: List every AI feature, model and vendor in the product, with the EU customers who use each | Output: AI inventory, owned by one person |
| 3 to 4 | Work: Decide role and tier per feature. Check intended purpose in code, docs, website and sales decks | Output: Role and tier register, with notes that counsel can review |
| 5 to 6 | Work: Fix Article 50 gaps: AI disclosure in chat, machine-readable marking of generated content, deepfake disclosure settings | Output: Shipped changes and a test per change |
| 7 to 8 | Work: Write the AI literacy record and a short internal AI policy. Check the roadmap against the prohibited list | Output: Training record, policy, sign-off |
| 9 to 10 | Work: For any Annex III candidate: start the risk register, the data sheet and the logging design | Output: Draft technical file outline and log schema |
| 11 to 12 | Work: Link incident handling to CERT-In, DPDP and Article 73. For a high-risk system, ask counsel about an EU authorised representative | Output: One incident runbook, representative decision |
| 13 | Work: Prepare a customer-facing statement: role, tier, dates, contacts. Date it and state what is not yet done | Output: A one-page statement your sales team can send |
Buyers in the EU will ask about this in security reviews. Keep the answers consistent with your AI security questionnaire answers. The EU's GDPR also applies when you process personal data of people in the EU, whatever the AI Act says. Our comparison of DPDP and GDPR for SaaS is a starting point for that work.
If you want help turning this into tenant settings, logs, evidence and a customer-ready statement, that is the work of our AI governance service. We build the system changes and the evidence. Your counsel signs off the legal position.
Sources: Regulation (EU) 2024/1689 (the AI Act), Articles 2, 3, 4, 5, 6, 9 to 15, 17, 22, 25, 49, 50, 72, 73 and 99 and Annex III, via artificialintelligenceact.eu and the European Commission's AI Act page (digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai). AI Omnibus (Regulation (EU) 2026/1744, OJ 24 July 2026), timeline per the Commission and the Future of Privacy Forum, "The AI Act Implementation Timeline: What Changes Under the AI Omnibus". EU AI Act Service Desk, Article 73. Last reviewed 20 November 2026.
FAQ
Does the EU AI Act apply to Indian companies?
It can. The Act covers providers that place AI systems on the EU market whether or not they are established in the EU. It also covers providers and deployers outside the EU where the output of the AI system is used in the EU. An Indian SaaS vendor with EU customers using its AI features is usually in scope as the provider.
Is my SaaS company a provider or a deployer under the EU AI Act?
A provider develops an AI system, or has one developed, and places it on the market under its own name or trademark. A deployer uses an AI system under its authority. An Indian SaaS vendor selling an AI feature to EU customers is usually the provider, and the EU customer is the deployer. Staff using a third-party chatbot inside your company make you a deployer for that tool.
When do the EU AI Act high-risk rules apply after the AI Omnibus?
The AI Omnibus moved the dates. Annex III high-risk systems, such as employment and education uses, now apply from 2 December 2027 instead of 2 August 2026. Annex I high-risk systems, which are AI in regulated products, apply from 2 August 2028 instead of 2 August 2027. Transparency rules under Article 50 have applied since 2 August 2026.
Do I need an EU authorised representative as an Indian vendor?
If you are a provider established outside the EU and your AI system is high-risk, Article 22 requires you to appoint an authorised representative in the EU, by written mandate, before you make the system available on the EU market. Systems that are not high-risk do not trigger that rule. Ask EU counsel to confirm how it applies to your product.
What are the fines under the EU AI Act?
Under the Act as adopted in 2024, fines reach up to 35 million euros or 7% of worldwide turnover for prohibited practices, 15 million euros or 3% for most other obligations, and 7.5 million euros or 1% for misleading information. The higher figure applies, except for SMEs, where the lower applies. Check with counsel whether the AI Omnibus changed any of these.
