M&A法務デューデリジェンスにおけるAI

In a typical acquisition, the seller uploads corporate records, contracts, employment documents, IP filings, litigation files and permits to a virtual data room hosted on a platform such as Datasite or Intralinks. The buyer's lawyers review the material to identify risks, check the statements the seller will make in the purchase agreement, and work out what must happen before closing. Contract review is the largest share of that work, and AI has been used for it longer than for most legal tasks. Kira Systems, now part of Litera, and Luminance are well-known extraction tools trained to find provisions such as assignment, change of control, termination, exclusivity, non-compete, most-favored-nation, indemnity caps and governing law. Newer generative AI tools can also summarize a clause's effect in plain language and answer questions across the whole data room. Change-of-control and assignment clauses need special attention. Whether a clause is triggered depends on both its wording and the deal structure. An asset purchase usually requires assigning contracts to the buyer. A stock purchase or merger may or may not trigger a clause, depending on whether it covers changes in ownership, mergers, or assignments that happen automatically by law. AI can find and classify the language, but deciding whether this deal needs consent is legal analysis. Diligence reports are usually built on structured extraction: a table listing each contract and its key terms, followed by a red-flag report summarizing issues by severity. AI speeds up both. It is often assumed to replace associate review, but it does not. Extraction misses unusual drafting, side letters, amendments that change the base agreement, and poor scans. Many teams have lawyers check every high-risk flag and a sample of the rest. The findings shape negotiation of the purchase agreement, the disclosure schedules and the closing conditions. Where the buyer uses representations and warranties insurance, they also inform the insurer's underwriting.