In recent years, companies specializing in artificial intelligence (AI) technologies have been increasingly coveted acquisition targets. With the AI field transforming our approaches to key issues – from climate change to cancer treatment – and pushing the boundaries of human capabilities through technologies such as self-driving cars and advanced space travel, it has the potential to dramatically alter life for humankind. Accordingly, it is no surprise that AI has emerged as a prime sector for M&A activity. As of October 2022, AI deals had surpassed $16.9 billion in aggregate value for the year and experts expect deal activity only to grow. Of particular interest are generative AI companies – those that produce algorithms able to create new content based on data inputs. Generative AI’s potential capability to revolutionize productivity has led to a considerable increase in investment activity and company valuations since 2020.
As active as the overall market has been, engaging the AI field does not come risk-free to the would-be acquirer. As the field has exploded in recent years, some of its familiar participants have warned of the unknowns and harms associated with the technology and its capabilities, as well as the companies developing it. Beyond the caution with which a traditional M&A transaction is treated, those seeking to purchase an AI company must approach the deal clear-eyed and deploy the resources and expertise to mitigate the associated risks.
Conducting due diligence on a target in the AI space may be considerably more difficult than on conventional technology companies. Buy-side advisors investigating a traditional technology company can readily review its proprietary software – usually at the heart of the transaction – for standard intellectual property concerns; for example, advisors will assess the company’s use of open source code and whether the underlying IP has been properly assigned to the company. The review of an AI company’s chief product can prove far more nebulous, as the company’s value is often derived from its datasets and proprietary models which absorb and analyze information.
Accordingly, when conducting diligence on an AI company, advisors are encouraged to expand and revise the scope of the review. They should review the company’s rights to its models, data, and “outputs” therefrom – a more difficult and nuanced investigation for the uninitiated. Buy-side advisors must draw on expertise in data ownership and privacy, as well as knowledge of the specifics of the AI field, to probe the company for risks to which a traditional technology target might not be as vulnerable. They must undertake a comprehensive review exceeding that of the traditional IP diligence investigation, both to assess the merits of a proposed transaction and to adjust the terms governing it appropriately.
Although this additional diligence may be cumbersome for both the target and buyer – and adds an additional layer of costs to the transaction – completing it can help the buyer better evaluate the AI capabilities of the target. Through this process a buyer can get a better understanding of various risks to which the target is exposed. For instance, many startup AI companies claim to have a fully automated system when in reality a lot of their processes are still manual; performing the additional diligence can help a buyer understand how far along the target’s AI technology actually is. In addition, through diligence, a buyer can determine whether the target’s AI technology is capable of handling real-world data or whether it has only been fine-tuned for a particular customer of the target. Diligence can also uncover risks associated with a target’s data management and compliance procedures, as well as potential regulatory risks to which the target may be exposed based on the type of data the target collects, processes, and stores.
To help guide your thinking, below is a non-exhaustive list of diligence questions that could be asked of acquisition targets working in the AI space during the evaluation process:
The standard form purchase agreement, including its representations and warranties, may not adequately address the risks involved in AI. But the representations and warranties contained in a definitive agreement will describe how the target company uses its AI assets and shift risks associated with those assets. Alongside the diligence review, these provisions (if drafted properly) will provide the buyer with the necessary information to understand more fully what it is acquiring.
As there are risks inherent and unique to the AI space not always encountered in traditional technology fields, the representations and warranties should be tailored specifically to the target company and its assets. The buyer’s legal advisors should be careful not to resort to generic, off-the-shelf language for a standard acquisition in the tech space. For instance, the buyer’s advisors should draft the representations such that risk is shifted to the seller with regards to the target’s rights to use the outputs associated with its AI models. By identifying such specific risks associated with AI and drafting the representations to allocate liability appropriately, a buyer investing in such AI technology can feel increasingly comfortable.
The definitions of the purchase agreement must also be crafted to the specifics of the field. The definitions pertaining to AI and the company’s products should be broad enough to capture the various techniques employed to create the AI. They should also adequately address the target’s specific niche in AI – a field of remarkable variety and expanse. As these definitions will govern the scope and applicability of the provisions contained in the agreement, it is crucial to tailor them to the specifics of the industry and target.
The active nature of the acquisition market for AI companies itself presents risks to both buyer and seller. Such activity has pressured companies to emphasize rapid growth, which can come at the cost of the long-term health and stability the company as well as its attractiveness as a target for acquisition. A potential buyer would do well to ensure the company has scaled appropriately and efficiently.
And, as significant as the targets’ AI products are to many of the recent acquisitions, these transactions are often driven instead by the desire to purchase talent—namely, the targets’ coveted AI researchers and engineers. This approach can be fraught with risks, from the bloating of the transaction value to downstream concerns of misalignment and retention. While any M&A transaction can raise questions of target compatibility, such risks are likely heightened by the particularities of the AI field.
Further, the unique, nascent nature of the field lends itself to a host of additional headaches for the would-be buyer—some of which perhaps escape the purview of a traditional due diligence review—from issues regarding the product’s quality and consistency to its future legal and ethical stature. While the AI field presents remarkable opportunity in both the narrow and broadest senses, buyers would do well to approach the transaction with the caution and sophistication the field warrants.