M&A transactions involving AI companies: representations and warranties

Introduction

The rapid advancement of artificial intelligence (AI) has made companies that develop and deploy novel, innovative AI technologies, as well as those that leverage AI to create significant competitive advantages in their products and services, highly attractive acquisition targets. These acquisitions present unique challenges since AI systems differ fundamentally from traditional technologies.

Unlike static systems, AI often involves self-learning capabilities that can change its behavior over time, leading to novel risks or manifesting risks in unpredictable ways. Additionally, the regulatory landscape for AI is complex and evolving, and the development and deployment of AI systems come with unique intricacies that require specialized attention.

While thorough due diligence is critical for identifying risks and opportunities — an area we addressed in a previous Update published on July 26, 2024 entitled “M&A transactions involving AI companies: due diligence considerations”, representations and warranties are equally important. To address the distinct risks of AI and to maximize the success of these acquisitions, traditional representations and warranties often need to be revisited and adapted.

This Update explores key considerations for crafting representations and warranties tailored to acquiring and investing in AI companies, building on insights shared by Osler partners Sam Ip (Technology) and Sophie Amyot (Corporate) in their recent webinar on M&A considerations for AI companies and large-language models.

AI representations and warranties

In M&A transactions, representations and warranties provided by the seller to the acquirer or investor serve as a risk allocation mechanism and form the basis for potential indemnification claims if losses arise from false or inaccurate representations and warranties. Representations and warranties must account for the unique characteristics of AI and the associated risks with its development, deployment and use. Unlike traditional software, where value often resides in the underlying software code, the “digital crown jewels” for AI companies are typically found in their models and datasets. These assets are not only critical to the company’s operations, but also present distinct challenges that require special consideration.

In 2025, the complexity of AI transactions continues to grow, most recently by advancements that are re-shaping even traditional software-as-a-service (SaaS) models and ushering in the era of autonomous AI systems, often referred to as “agentic AI.”  This shift exacerbates the complexity of AI transactions, as agentic AI systems are capable of independent decision-making and self-learning, introducing new layers of risk, such as unpredictability in their evolution and the potential for unintended outcomes. These trends underscore the need for a tailored approach to representations and warranties that capture these novel risks and liabilities.

Key components of AI-tailored representations and warranties include:

• Establish key AI definitions: clearly define AI assets, including models, training data, algorithms, and system components, to ensure precision and avoid ambiguity in transaction documents. These definitions form the foundation for representations and warranties and must strike the balance between comprehensive and over-inclusive (i.e., capturing non-AI assets). Avoiding this common pitfall is critical to reducing friction during negotiations.
• Disclosure of AI-related IP and related use cases: ensure comprehensive disclosure of key AI assets, such as models, datasets, and related intellectual property. Focus on use cases, as risks can vary significantly depending on how AI technology is applied. For example, AI used in health applications will have vastly different regulatory and liability concerns than AI used in entertainment.
• Ownership of AI: expand traditional ownership representations to encompass not only software, but also AI-related models, datasets, and algorithms. Ensure the target has clear title to these assets and has secured necessary rights to use and commercialize them, with a particular emphasis on third-party data licensing agreements or datasets created through partnerships.
• Data use rights and data quality: augment standard data use representations to include assurances regarding the quality, accuracy, and legality of the data used for training AI systems. Address potential issues related to licensing, privacy, regulatory compliance, and provenance of training data to mitigate downstream risk and liability.
• AI-specific risk:  incorporate representations addressing AI-specific risks, such as transparency, bias, accuracy, and explainability. These representations will need to be context-specific. For example, explainability may be critical for regulated sectors such as finance or healthcare where accountability is paramount.
• Compliance with evolving AI standards: consider inclusion of compliance with leading frameworks (e.g.,  NIST AI RMF or ISO/IEC 42001). This ensures target adheres to industry best practices.
• Open source risk: address the growing reliance on open source AI models by including representations that ensure compliance with applicable open source licenses. This mitigates risks associated with copyleft obligations or IP contamination, and any conflicts with proprietary and other open source licenses.
• AI governance:  ensure representations and warranties cover the target’s AI governance practices, including adherence to ethical AI guidelines, conducting regular AI impact assessments, and implementing documented human oversight processes for decision-making. Robust governance often reduces risks related to unethical AI deployment.

Since the foregoing will not apply universally, thoughtful consideration is required to tailor representations and warranties to the specific circumstances of each transaction, ensuring they adequately address the risks and value drivers associated with the target’s AI assets.

Risk mitigation: closing and post-closing

Traditional risk allocation tools, such as escrows, holdbacks, and indemnities, remain relevant for giving effect to AI-tailored representations and warranties in acquisition agreements. In some cases, it may be appropriate to introduce separate liability caps and survival periods for AI-related representations and warranties, as well as targeted indemnities to address AI-related risks identified in due diligence. Additionally, representation and warranty insurance is emerging as a valuable tool in AI transactions, offering customizable coverage for risks that are difficult to quantify or resolve during negotiations.

Closing conditions also can be leveraged to mitigate risks associated with AI technologies by requiring the target to resolve identified deficiencies — such as gaps in ownership, data use rights, privacy compliance, or regulatory adherence. 

Post-closing, acquirers should conduct follow-up reviews to confirm compliance, verify data use rights, and monitor the performance of AI systems to identify and address any residual risks. Legal advisors can play a critical role in conducting these reviews, offering strategic insights to ensure acquirers maximize the value of their newly-acquired AI assets and mitigate ongoing risks effectively.

Incorporating these measures in both the closing and post-closing stages of an AI transaction provides acquirers and investors some levers to mitigate related risks and to enhance the return on their investment.

Conclusion

Thorough due diligence, clear definitions, and carefully crafted representations and warranties, are essential for acquirers and investors of AI companies. Unlike traditional software, AI assets present unique complexities, requiring a tailored approach to address their distinct risks and nuances. For AI companies preparing for acquisition, proactively planning for the considerations described in this Update will help meet investor and acquirer expectations and navigate the transaction process more effectively.

By addressing AI-specific risks head-on, both acquirers and AI companies can align expectations, mitigate potential challenges, and set the foundation for a successful and efficient transaction. As AI continues to reshape industries and drive innovation, staying ahead of evolving best practices in AI acquisitions will be critical for companies and acquirers looking to capitalize on the transformative potential of this technology.