91视频

Private equity and venture capital investment in 91视频 artificial intelligence companies has grown sharply in the past decade, driven by the rapid deployment of AI across sectors, from fintech and healthtech to logistics and enterprise software. Unlike a typical software company where due diligence focuses on source code ownership, vendor contracts and customer concentration, an AI company derives its core value from the interaction of three distinct layers: the underlying training data; the algorithms and model architecture that process that data; and the outputs those models generate. Each layer carries its own legal, regulatory and commercial risk profile.

While recent reports suggest intent to introduce specific legislation governing AI, India does not yet have a dedicated statute in this regard. In its absence, the applicable legal framework is fragmented across data protection, intellectual property (specifically copyright protection), and sector-specific statutes, each of which may interact with the target’s business in ways that are not always immediately apparent.

Data provenance, quality and consent

Data is the foundational input for most AI systems and its provenance, quality and legal basis for use are among the most consequential diligence questions. Key considerations include the sources of training data and how they have been obtained – whether proprietary, licensed from third parties, scraped from public sources, or generated synthetically – as well as the terms governing their storage and use.

Where training datasets contain personal data, compliance with applicable data privacy laws is critical, especially the procurement of specific consent for such use. Consent obtained for one purpose does not automatically extend to training an AI model, and data collected under legacy privacy policies may not satisfy current legal requirements.

Licences, IP stack and ownership

Data use restrictions embedded in third-party licences, data sharing agreements, or customer and vendor contracts must also be assessed to identify any material limits on the target’s ability to commercialise its products and the corresponding implications for its business model.

The IP stack of an AI company typically comprises model architecture and weights, the inference pipeline, outputs generated by the system and, ideally, training datasets. Each element requires separate analysis depending on the nature of the target’s business.

From an ownership perspective, investors must verify a clean chain of title from the individuals who developed the algorithms and models to the target company. Open-source software introduces a distinct set of risks often underestimated in AI diligence.

Open-source, outputs, talent and governance

Many foundational AI frameworks are typically accessed under permissive or copyleft licences, the latter requiring further use to be subject to similar terms. Accordingly, software incorporating such open-source code or based on open standards, even if touted as unique, may need to be released under the same open-source terms, potentially destroying proprietary value.

Autonomously generated AI outputs are not expressly addressed under the Copyright Act, 1957. Their position in India remains unsettled and evolving pending the conclusion of judicial discourse.

A key consideration in AI companies’ diligence, particularly at an early stage, is the personnel – data scientists, engineers and research leads whose expertise is central to the target’s business. Recent AI startup acquisitions in the US use an “acqui-hire” structure, combining the hiring of talent with the simultaneous licensing of AI technology from the startup, rather than a standard share-based acquisition.

It is also important to check the AI governance policies in place, including measures to prevent algorithmic bias, discrimination and model drift. The existence of periodic audit trails and IS/ISO/IEC 42001:2023 certification are indicators of institutional maturity.

AI diligence requires different lens

While the fundamental mechanics of a PE/VC diligence exercise remain constant, the diligence of an AI company demands a materially different lens from the one that is applied to a conventional technology or software business.

Key considerations of data provenance and consent, IP ownership and open-source hygiene, as well as key personnel retention must be examined with the same rigour as financial and commercial diligence metrics. Investors who build this capability, and who insist on AI-specific representations, warranties and indemnities in transaction documents, will be better positioned to protect value and manage the unique risks presented by AI companies.

Vandana Pai is a senior partner and Shreya Sreesankar is a senior associate at Bharucha & Partners