How buyers assess AI and LLM features in due diligence

What investors now ask about AI and language models in product diligence—and how to prepare defensible answers.

AI features moved from novelty to standard diligence scrutiny. Buyers typically probe five areas.

Capability vs claim — Does the product do what marketing and the IM imply?
Data — Training/fine-tuning data rights, PII, retention, cross-border transfer.
Cost — Unit economics of inference, rate limits, fallback when models change.
Security — Prompt injection, output handling, access to sensitive context.
Vendor lock-in — Dependence on a single model API or hosted stack.

Sell-side vendor DD should document honest answers and remediation where gaps exist—before a buyer’s specialist team reframes them as price chips.