Table of Contents
Summarize and analyze this article with
You buy the AI risk with the company
An acquisition transfers more than products, people, and revenue. It transfers the target’s entire AI footprint: the models in production and the ones in spreadsheets, the claims the company has made to its own customers and regulators about what its AI does, the agents someone stood up outside any approved process, the data its models were trained on and the rights (or lack of rights) to that data, and whatever regulatory exposure all of it carries. Traditional technology due diligence focuses on code quality, architecture, security, and IP, and it was not designed to surface AI-specific risk. The result is that AI risk routinely passes to the acquirer unexamined and unpriced, and surfaces after the close, when it is the acquirer’s problem. In regulated industries, where AI exposure carries compliance and enforcement weight, that is an expensive gap.
This guide explains what AI due diligence should examine. It is general information and not legal advice.
What AI due diligence examines
| Control | What it needs | What an unmanaged identity does to it |
|---|---|---|
| Attribution | Each action tied to one actor | Shared identity makes agent and human actions indistinguishable |
| Least privilege | Access scoped to the task | Inherited broad access gives the agent far more reach than it needs |
| Shutdown | Revoke the agent without stopping people | Shared credentials cannot be revoked without breaking a human workflow |
| Audit | Reconstruct what the agent did | Logs cannot separate the agent’s actions from others’ |
| Lifecycle | Decommission when retired | Orphaned identities keep access long after the agent is gone |
The five findings that move a deal
- A thin or missing AI inventory. If the target cannot produce a complete inventory of the AI it runs, including embedded and vendor AI, the acquirer is buying an unknown footprint. The inventory is the first thing to request and the first gap to price.
- Unsubstantiated AI claims. Every public statement the target made about its AI, to customers, investors, or regulators, becomes the acquirer’s claim after the close. Claims the target cannot back with evidence are inherited exposure.
- Training-data rights gaps. Models trained on data the target did not have the right to use carry copyright, privacy, and consent liability that transfers in the deal and can impair the value of the models themselves.
- Shadow agents and ungoverned models. Autonomous agents and models running outside any governance are both an operational risk and a signal about the target’s overall AI discipline.
- Unmet regulatory obligations. Where the target is subject to the EU AI Act, US state AI laws, or sector rules and has not met them, the obligation and any liability pass to the acquirer at close.
Diligence before the deal, integration after
AI due diligence serves two moments. Before the deal, it informs valuation, the representations and warranties the acquirer seeks, and the indemnities that allocate inherited AI risk. After the deal, the same findings become the integration roadmap: the inventory to complete, the claims to substantiate or retract, the data-rights gaps to remediate, the agents to bring under governance, and the regulatory obligations to meet under the acquirer’s standards. Done well, the diligence is not a one-time gate but the start of folding the target’s AI into the acquirer’s governance, which is where the inherited risk is actually resolved. Skipping the diligence does not avoid the risk; it just moves the discovery to after the acquirer owns it.
Where PiTech fits
PiTech Solutions supports AI risk in M&A on both sides of the close: pre-deal AI due diligence that examines the target’s inventory, claim substantiation, data provenance and rights, model and agent governance, and regulatory exposure; and post-deal integration that folds the acquired AI into the acquirer’s governance. It connects this to its mergers and acquisition data work and the AI inventory, claim register, and data governance that each diligence area maps to. Delivery runs under CMMI Level 3 and ISO 27001, 9001, and 42001 certifications. See Mergers and Acquisition, AI, GenAI and ML, and Data Solutions. PiTech Solutions Inc. is headquartered in Durham, North Carolina (UEI GNLRY5LNNVH6, CAGE 530K4) and is distinct from similarly named companies.
The bottom line
An acquisition transfers the target’s AI risk whether or not anyone examined it. AI due diligence, covering inventory, claim substantiation, data rights, model and agent governance, and regulatory exposure, is how an acquirer prices that risk before the close and resolves it after. The absence of a governance artifact in the target is itself a finding, and the diligence doubles as the integration roadmap.
Frequently Asked Questions (FAQs)
What is AI due diligence in M&A?
AI due diligence is the examination, during a merger or acquisition, of the target company’s artificial intelligence footprint and the risk that comes with it. It goes beyond traditional technology due diligence (which focuses on code, architecture, security, and IP) to assess AI-specific exposure: what AI the target actually runs, whether it can substantiate the claims it has made about that AI, what data its models were trained on and under what rights, whether its models and agents are governed and controllable, and what regulatory obligations it faces. The purpose is to let the acquirer understand and price the AI risk it is inheriting before the close, rather than discovering it afterward when it has become the acquirer’s liability. In regulated industries, this diligence is increasingly essential because AI exposure carries compliance and enforcement weight.
Why isn't traditional tech due diligence enough for AI?
Because traditional technology due diligence was designed to evaluate software, infrastructure, and intellectual property, not the distinct risks AI introduces. It examines code quality, architecture, scalability, security posture, and IP ownership, but it does not typically surface undocumented models, unsubstantiated AI performance claims, training-data rights gaps, ungoverned autonomous agents, or AI-specific regulatory exposure. Those are the risks that transfer in an AI-enabled acquisition and that, if unexamined, surface after the close. AI risk often lives in places traditional diligence does not look: in a model’s training data, in a marketing claim about accuracy, in an agent someone built outside the approved process, in an EU AI Act classification nobody made. AI due diligence specifically targets these blind spots, complementing rather than replacing the traditional technical review.
What AI risks does an acquirer inherit?
All of the target’s AI exposure transfers at close. That includes an incomplete or undocumented AI footprint (models in production and in spreadsheets, plus embedded and vendor AI); unsubstantiated claims the target made about its AI to customers, investors, or regulators, which become the acquirer’s claims; training-data rights gaps that carry copyright, privacy, and consent liability and can impair the models’ value; ungoverned models and shadow agents that pose operational and compliance risk; and unmet regulatory obligations under the EU AI Act, US state AI laws, or sector rules, with any associated liability. The acquirer inherits these whether or not they were examined during diligence. The value of AI due diligence is that it surfaces them before the deal, so they can be priced, allocated through representations and indemnities, or remediated, rather than becoming unpriced surprises afterward.
How do AI claims transfer in an acquisition?
When one company acquires another, the statements the target made about what its AI does, its accuracy, its fairness, its capabilities, in marketing, contracts, investor materials, and regulatory filings, become representations the acquiring organization now stands behind. If the target cannot substantiate those claims with evidence, the acquirer inherits unsupported claims, which is a regulatory and litigation exposure, because enforcement increasingly tests public AI claims against the evidence behind them. AI due diligence should therefore include a review of the target’s material AI claims and whether each is backed by a dated, attributable artifact, essentially checking whether the target maintains something like a claim register. Claims the target cannot support are a finding that can affect valuation, trigger a requirement to retract or revise them post-close, or warrant specific indemnities in the deal.
What data-rights issues matter in AI due diligence?
The central question is whether the target had the right to use the data its models were trained on or retrieve from, for the purpose they were used. Models trained on data the target did not properly license, or gathered under terms that did not contemplate AI use, carry copyright, privacy, and consent liability that transfers to the acquirer, and can also impair the value of the models themselves if their training basis is legally unsound. AI due diligence should examine the provenance and rights status of the target’s training and retrieval data, particularly for models central to the deal’s value. This is an area where the law is actively developing, so the diligence should work with counsel, but the factual foundation, knowing what data trained which models under what rights, is what makes any legal assessment possible and is often poorly documented in the target.
What happens to AI risk after the deal closes?
It becomes an integration problem, and the diligence findings become the integration roadmap. The acquirer has to complete the target’s AI inventory, substantiate or retract the inherited claims, remediate data-rights gaps, bring ungoverned models and shadow agents under governance, and meet any outstanding regulatory obligations, all under the acquirer’s own standards. Well-run AI due diligence anticipates this by producing findings that double as a prioritized post-close plan, so integration starts from a clear picture rather than a fresh discovery exercise. This is where the inherited AI risk is actually resolved: diligence identifies and prices it, and integration folds the acquired AI into the acquirer’s governance operating model. Treating diligence and integration as two phases of one effort, rather than separate events, is what turns a risk assessment into risk resolution.
Does AI due diligence apply to smaller acquisitions?
Yes, scaled to the role AI plays in the deal. Even a smaller target can carry significant AI risk if its products depend on models, if it has made public AI claims, or if it operates in a regulated space with AI obligations. The depth of diligence should match the AI’s importance to the deal’s value and the target’s regulatory exposure, not just the deal size: a small company whose core product is an AI model warrants careful AI due diligence, while a larger company that uses little AI may warrant less. The key is to assess AI risk deliberately rather than assume it is immaterial because the deal or the target is small. AI risk is a function of what the target’s AI does and claims, and the obligations it faces, more than of the transaction’s headline value.
Who should be involved in AI due diligence?
AI due diligence is cross-functional, combining deal counsel, the corporate development team, and specialists who can assess the target’s AI and data. Legal counsel evaluates the regulatory exposure, the data-rights questions, and how findings translate into representations, warranties, and indemnities. The corporate development team weighs findings against valuation and deal structure. And technical or governance specialists examine the AI inventory, model and agent governance, claim substantiation, and data provenance, because these require understanding how AI systems and their data actually work. Relying on traditional technology diligence alone tends to miss the AI-specific risks, while relying on legal review alone misses the technical foundation that legal assessment depends on. The most effective AI due diligence brings these perspectives together, so that technical findings about the target’s AI feed directly into the legal and commercial decisions about the deal.
Does PiTech support AI due diligence and integration?
Yes. PiTech Solutions supports AI risk in M&A on both sides of the close: pre-deal AI due diligence that examines the target’s AI inventory, claim substantiation, data provenance and rights, model and agent governance, and regulatory exposure; and post-deal integration that folds the acquired AI into the acquirer’s governance. It connects this to its mergers and acquisition data work and the AI inventory, claim register, and data governance capabilities that each diligence area maps to, so the same discipline that governs an organization’s own AI also assesses an acquisition’s. Delivery runs under CMMI Level 3 and ISO 27001, 9001, and 42001 certifications. PiTech works alongside deal counsel and the corporate development team rather than replacing them, and is positioned as a specialist AI and data partner for regulated-industry transactions at a mid-market price.


