Key Points

• M&A in the AI sector is likely to increase as the technology and the AI ecosystem mature.
• Deals in the AI sector pose unique legal and regulatory challenges, including potential antitrust and/or national security reviews.
• Comprehensive due diligence is crucial, and acquirers are beginning to require tailored representations and warranties to address risks that are particular to AI.
• Deal terms are also evolving to address other AI-specific issues, including material adverse effect definitions, interim operating covenants and recourse mechanisms such as indemnities and representation and warranty insurance.

The accelerated development of artificial intelligence (AI) has shown the transformative potential of the technology across industries, making it an integral part of strategic planning for market participants, from technology giants to venture capitalists. Unsurprisingly, AI technology has drawn an enormous wave of investment.

As with other technology booms, companies with market-leading solutions and a clear path to growth and profitability will emerge as leaders, while those unable to demonstrate tangible value or competitive advantage are likely to struggle.

We anticipate a surge in M&A throughout the AI ecosystem as the AI market develops, posing a range of new legal issues and in some cases requiring AI-specific approaches.

In particular, as AI companies seek to consolidate to achieve synergies and strengthen their market positions, regulators are likely to raise antitrust concerns. There has already been intensifying scrutiny of antitrust authorities over the existing landscape, where leading technology companies have raced to partner with promising AI startups. Additionally, the strategic importance of AI technologies to national security is also likely to increase government oversight of cross-border transactions.

These factors will complicate M&A transactions in the sector, necessitating a sophisticated approach to any such transactions.

Given the complexity and nascent nature of many AI technologies, acquirers should undertake comprehensive due diligence to accurately assess the value and potential risks associated with their targets.

For certain targets, such as companies providing data centers and AI infrastructure, due diligence will likely remain largely the same as those applicable to technology companies more generally, though sometimes with an additional focus on such companies’ AI-related capabilities (e.g., for a data center target, an acquirer may want to confirm whether such target has access to the power generation and specialized chips needed for AI).

But for other segments of the AI ecosystem, especially those heavily reliant on AI systems and their applications, due diligence should be adapted accordingly.

Increasingly, we are seeing targeted inquiries regarding the unique characteristics and potential risks associated with AI technologies that are critical drivers of value or areas of enhanced regulatory scrutiny.

In light of potential risks uncovered in due diligence and regulatory uncertainties, market participants are focusing more and more on tailored deal terms in the transaction agreement to address unique issues raised by AI. Parties are adding certain deal terms to manage these concerns and uncertainties.

Representations and Warranties

In most cases, the representations and warranties in the acquisition of AI companies do not deviate significantly from those typical in the acquisition of other high-tech companies. Many of the significant risks associated with AI companies, such as those related to intellectual property (IP) rights, information technology (IT), data privacy and cybersecurity, are similar to those other high-tech companies encounter.

As such, it is often sufficient for an acquirer to rely on standard representations in a transaction agreement (e.g., regarding the absence of IP infringement or cyber breaches).

However, acquirers are increasingly pushing for separate AI-specific representations in situations where a particular feature or aspect of a target’s AI technology is crucial to the valuation of the business or involves unique risks. Tailored representations are also being used in situations where the existing law is unsettled, most significantly in the case of generative AI.

For instance, current legal standards are unclear about whether it is permissible to use web-scraped data, which often include copyrighted works and personal data, to train an AI model. If a target engages in this practice, relying on general noninfringement or compliance-with-laws representations may not provide sufficient assurance. Therefore, an acquirer may require a specific representation that the AI model was trained only with permissioned data.

Similarly, where the law is nascent or evolving, such as with respect to bias, explainability and trustworthiness of AI models, an acquirer may require confirmation that the target’s AI systems are compatible with the acquirer’s overall risk tolerance or corporate philosophy regarding such matters.

Definition of ‘Material Adverse Effect’

The definition of “material adverse effect” is often a key negotiating point in transaction agreements, as it may be used to qualify certain of a target’s representations, and the absence of a material adverse effect is a standard closing condition. The definition usually carves out many effects that are not within a party’s control, such as changes in general political or economic conditions, wars, natural disasters and changes in law or accounting principles.

Given the evolving regulatory landscape, sellers of an AI business may try to specify that the changes-in-law exception includes updates to laws concerning AI technologies.

Interim Operating Covenants

Interim operating covenants in a transaction agreement are intended to preserve the value of the target and ensure that the acquirer receives the business in substantially the same condition as at the time the agreement was signed. Acquirers may negotiate to include AI-specific actions that a target cannot take without the acquirer’s consent, given the rapid development in the AI sector and the changing regulatory environment.

Examples of such prohibitions on a target’s actions between signing and closing include that the target may not:

• Materially change the nature of training data, terms of use governing the target’s AI-powered products or services, or relevant compliance policies and procedures.
• Implement new AI practices, policies and procedures.
• Acquire or in-license AI-related assets or onboard a new AI vendor.

Recourse

As discussed above, an acquirer of an AI business can generally rely on standard IP, IT, data privacy, cybersecurity and compliance-with-laws representations, which cover a broad range of common risks associated with high-tech companies.

Depending on the nature of the target’s business, the acquirer may consider treating some of these representations — and any additional AI-specific representations included in the agreement — as “fundamental representations” or a different category with longer survival periods and higher caps than “general representations,” to ensure adequate protection and risk mitigation.

In addition, if any specific AI-related issues are uncovered in due diligence, the acquirer may seek to include a special indemnity, which would often be subject to negotiated limitations (e.g., baskets, caps, procedures).

As an alternative or supplement to indemnification, an acquirer may seek representation and warranty insurance (RWI) to protect it from financial losses arising from undisclosed or inaccurately presented information about the target.

To date, there have not been significant differences in underwriting practices and standards between transactions in the AI sector and those in the broader technology industry.

However, we anticipate increased underwriting scrutiny from RWI carriers, as more novel, AI-specific representations like those described above are included in transaction agreements. Policy coverage and exclusions may evolve to address the unique risks associated with AI. RWI carriers might also leverage AI tools themselves to independently assess an acquirer’s due diligence findings.

[View source.]