Reproduced with permission. Originally published January 10, 2024, “Q&A: How generative AI will play a role in dealmaking,” Westlaw Today with permission of Thomson Reuters.

Generative artificial intelligence is changing the business world and impacting how mergers and acquisition dealmakers operate.

Wai Choy, a partner at Proskauer Rose LLP, discusses the role that generative AI will play in mergers and acquisitions, how it will help dealmakers and potential issues that may arise during transactions.

Thomson Reuters: What do mergers and acquisitions dealmakers need to know about generative AI?

Wai Choy: It's critical to recognize that not all GAI platforms are created equal — especially as the number of GAI options increase and use cases proliferate. In addition to differences in technical aspects, the governing legal terms and conditions can vary greatly from one to the next. Determining the implications of GAI use for a business is a nuanced analysis that requires an understanding of which, and for what purposes, GAI tools are being used, as well as the nature of the data sets used in training the GAI model and the prompts that users submit.

GAI's ability to create new content in response to a user's prompt, based on what the GAI has learned from the content it was trained with, holds tremendous potential to reduce costs, increase efficiency, improve quality, deliver valuable insights and assist creativity. At the same time, GAI presents important legal, practical and commercial risks that need to be carefully managed. It is essential to focus on GAI issues in the M&A due diligence process and in negotiating the purchase or merger agreement, whether the target is a GAI service provider or user.

The use of GAI by any enterprise raises various confidentiality, intellectual property, accuracy, bias, privacy, cybersecurity and legal compliance considerations, which are important for acquirers and target companies to evaluate.

TR: Will generative AI use raise concerns for a business being targeted for a merger or acquisition?

WC: GAI does raise potential concerns for target companies (both GAI users and GAI service providers), which acquirers will analyze. As a result, target companies should ensure that they have a clearly defined GAI policy and technical measures to facilitate and monitor compliance, conduct adequate training so that their workforce understands how to safely and productively use GAI (e.g., what to avoid inputting into GAI tools and how to responsibly use GAI outputs), and negotiate agreements with GAI providers for enterprise versions of GAI solutions that include appropriate protections.

In addition to other specific risks that apply to particular businesses and use cases, the following five GAI considerations are generally relevant:

1. Confidentiality: The GAI provider may or may not be contractually obligated to treat user inputs and outputs confidentially. If not, content that is shared with or produced by certain GAI tools may be used to train their models, which could result in leakage of proprietary information and loss of trade secret protection. In addition, submitting a third party's confidential information to a GAI tool could breach confidentiality obligations owed to that third party.
2. Intellectual property: The governing legal terms and conditions may grant the GAI service provider rights to use the user's inputs to train GAI models, and submitting third-party licensed content to a GAI solution could violate the scope of that license. The extent to which IP rights in outputs of GAI tools can be owned is currently unclear in the U.S., so care should be taken when using GAI to develop content in which a company expects to have protectable proprietary rights. Where GAI is used to generate source code, outputs could include source code that is subject to an open source license (which may not be apparent and could impose obligations or undermine the company's proprietary rights). Also, GAI tools can infringe third-party IP rights, both in the training of the model and in the outputs they create. Recently, some GAI service providers have begun to offer indemnification of enterprise customers for certain copyright infringement claims arising from the customers' use of GAI outputs, but not all GAI providers do, and those indemnifications are subject to exceptions and caveats.
3. Accuracy and bias: GAI models can produce biased or inaccurate responses, such as "hallucinations" that read coherently but are false or misleading. This can be caused, for example, by inaccurate, incomplete or out-of-date training data.
4. Privacy: Submitting personal data into a GAI model, or using personal data in GAI outputs, could violate privacy laws and the rights of the individuals to which the personal data relates and could also breach the terms and conditions of the GAI provider. Relatedly, the use of GAI in employment decision-making is one hot-button area that has drawn focus from state regulators.
5. Cybersecurity: Although GAI solutions can be either cloud-based or locally deployed on a customer's systems, both raise cybersecurity issues. The rapidly evolving nature of the technology means that there may be unidentified vulnerabilities, which can be exploited by attackers. There have already been a number of GAI security breaches. In addition, the way in which GAI works raises unique cybersecurity risks, such as "model inversion" attacks, through which an attacker may be able to use outputs to derive the original training data.

GAI service providers should ensure that their terms of service are calibrated to adequately protect the company and their customers, and that their data gathering and model training practices are structured to be compliant with intellectual property and other laws.

TR: What areas of the M&A process will generative AI help and how will it help? How will attorneys working on M&A deals use generative AI?

WC: While human review, expertise and judgment will continue to be essential, as the capabilities of GAI continue to expand rapidly, there are ways in which GAI can assist in the M&A process, such as:

• Processing unstructured data, like news articles, industry reports, public filings and social media posts, to identify and assess potential targets and understand market trends.
• Generating drafts of nondisclosure agreements, letters of intent, term sheets, due diligence request lists, purchase or merger agreements and other legal documents (which could leverage precedent documents from prior deals).
• Facilitating the comparison of the current deal terms with positions taken in prior deals and suggesting modifications to contractual provisions to align with the corresponding language in precedent agreements.
• Automating due diligence review, summarization and data extraction.
• Building financial models and predictive models.

TR: What potential legal issues could we see with dealmakers using generative AI?

WC: The legal terms and conditions of some GAI solutions (e.g., free, consumer-targeted or non-enterprise versions) can allow GAI providers to use all user inputs to train their models, meaning that if confidential information (such as target company data room documents) is submitted to such a GAI solution, it may result in proprietary information of the target company being disclosed in response to a third party's prompt and could be a breach of the confidentiality agreement governing the negotiation of the M&A deal.

In addition, even where an acquirer has an enterprise agreement in place with a GAI provider that includes appropriate confidentiality, cybersecurity and other protections, and the acquirer's confidentiality agreement with the target company permits the acquirer to input the target company's documents and data into a GAI solution, it is possible that a data breach could result in the exposure of those materials, which could result in legal liability to the acquirer.

TR: How will generative AI change deals in M&A? What will it do?

WC: The efficiencies offered by GAI can enable faster deal execution by shortening due diligence and definitive agreement negotiation timelines, which could provide acquirers that strategically leverage GAI with a competitive advantage, especially in a dynamic M&A environment.

TR: Will businesses that use generative AI be a target for acquirers?

WC: As the use cases and sophistication of GAI tools continues to develop, every business is likely to leverage GAI in at least some ways, meaning that GAI considerations will always need to be taken into account. We have seen immense interest in investing in and acquiring GAI service providers, as well as other companies that have built their own GAI models for internal use. In our deal flow we are already seeing this trend continue in 2024.