AI due diligence applications require close human oversight

Artificial intelligence is becoming a value-enhancing tool in private equity transactions. Mergers and acquisitions require careful attention to evaluating opportunities, risks, and compliance. The introduction of AI has revolutionized this landscape, increasing speed, accuracy, and cost efficiency.

However, relying entirely on AI diligence or failing to thoroughly investigate the output it generates can result in common post-closing claims such as fraud, breach of representation, and diminished business value.

AI tools can review thousands of documents, provide concise summaries of materials, analyze complex financial metrics, and operate in virtual data rooms. These positive attributes are contributing to the increasing use of AI in due diligence summarizing a company’s financial health, pending litigation, and extracting M&A information.

AI is not new to legal practice or exercises like due diligence. Predictive coding and technology-assisted reviews are useful precedents. Technology-assisted review increases efficiency and identification of relevant documents during the discovery process.

However, while AI can meaningfully assist with these tasks, it cannot replace the judgment of lawyers in assessing risk, data bias, pending litigation, or identifying potential legal minefields.

In recent years, courts have sanctioned lawyers for employing AI in legal research and for producing false citations and inaccurate precedents to support their positions. Over-reliance on AI tools creates risks when they generate “hallucinatory” data that leads to inaccurate valuations of the target company, increasing the likelihood of post-closing litigation.

For example, if a buyer discovers after closing that the seller has materially overstated the company’s financial performance because the AI did not accurately evaluate the numbers, this could lead to a material misstatement claim. To avoid inaccuracies, it is important for attorneys to understand the AI platform being used and its operational scope and limitations before relying on the output.

Companies and law firms will need to scrutinize the selection of AI vendors, conduct training, increase scrutiny of AI-generated results, and establish AI protocols to reduce the possibility of hallucinations.

The Securities and Exchange Commission is ramping up its crackdown on claims of “AI washing” by scrutinizing companies’ public marketing materials, regulatory filings, and social media posts about AI. AI washing occurs when a company attempts to attract investment by mischaracterizing its AI capabilities or making false claims about its use of AI tools. AI washing increases the risk of securities class action lawsuits being brought against companies that make such misstatements.

A crackdown on AI cleaning will need to coexist with the Trump administration’s focus on championing AI development. The administration suspended previous consent orders against AI companies for their products, issued an executive order in January 2025 requiring federal agencies to avoid stifling AI innovation, and directed enforcement reductions as part of a subsequent July 2025 AI Action Plan.

Lawyers should use data-backed evidence and documentation about a company’s technology to identify and confirm a company’s claimed AI capabilities. The due diligence component should focus on the seller’s use of AI in its operations to ensure accurate reporting.

Increased scrutiny of the use of AI will continue to extend to due diligence in the transactional arena. For example, assessing and evaluating litigation risk is an important part of due diligence and requires knowledge and background regarding pending litigation and the potential financial impact on the business.

Considering the litigation risks underlying a proposed transaction must include understanding the nature of the company’s business, potential litigation avenues, defense costs, and verdict risk. You should also consider the influence of sociopolitical issues that influence your client’s case assessment, risk management, or attitude.

AI cannot mirror the core functions performed by lawyers. Formal Opinion 512 from the American Bar Association notes that generative AI tools “lack the ability to understand the meaning or assess the context of the text they generate.” Contextual understanding is limited, as only lawyers can recognize subtle risks and subtle ambiguities.

AI may not recognize the impact of litigious succession, jurisdictional nuances, attorney reputation, impact on product formulation changes, issuance of warnings, competitive significance of non-compete clauses, or under-reporting of disclosed litigation.

Claims of fraud, misrepresentation, breach of warranty, and intellectual property issues can result from AI-powered tools lacking sufficient oversight and validation.

Certain AI platforms are intentionally general-purpose or focused on early-stage contract platforms and are trained using standard commercial contracts (non-disclosure agreements, service and vendor agreements). These are only useful when identifying missing standard clauses and reviewing large volumes of contracts and amendments. Such platforms cannot detect the nuances of highly specialized or niche contract clauses and subtle legal risks.

Verifying the authenticity of data is a key requirement as it ensures the accuracy, consistency and reliability of trading decisions.

Post-closing claims typically arise from pre-closing issues rather than unforeseen events. Claims arise from transaction structures that fail to predict liability. Buyers, sellers, and lawyers will need to reframe how they view and use AI in due diligence and avoid confusing fluency with accuracy.

AI-powered analytics can enhance and support the diligence process as an insightful amplifier for critical analysis, but not for decision-making.

This article does not necessarily reflect the opinion of Bloomberg Law, Bloomberg Tax, Bloomberg Government, publisher Bloomberg Industry Group, Inc., or its owners.