Before You Use AI: A Checklist for Protecting Confidential and Proprietary Information - Full Version

AI tools have shifted from optional add-ons to essential infrastructure. AI can take notes at meetings,

analyze and summarize data and reports, search and synthesize technical information, and prepare

investor and board decks, to name just a few. And DHL supports the use of AI throughout a wide range of

company processes, including intellectual property-related work.

But AI tools, without the appropriate legal guard rails, can jeopardize your IP. Letting AI capture

sensitive information without the right Terms & Conditions can lead to unintended disclosures, which can end up with competitors, forfeit patent rights, strip trade-secret protection, and waive attorney-client privilege.

The Terms & Conditions are key. Run the checklist below on an AI provider’s Terms & Conditions,

and the answers will tell you if this AI tool provider has the necessary legal safeguards to ensure your

confidential information and other valuable IP is secure.

AI Terms & Conditions Checklist¹ :

☐ You own all recordings, transcripts, and summaries.

☐ You control retention and permanent deletion.

☐ Your data is not used to train AI models.

☐ The vendor takes reasonable steps to maintain confidentiality.

☐ Third-party access is limited to a need-to-know basis and subject to Terms & Conditions atleast as restrictive as your own.

If you can’t confirm these points in the Terms & Conditions, look for another AI provider.

Here are 3 providers whose Enterprise Terms & Conditions passed our test, at least as of this publication date (but confirm for yourself before implementing them):

Anthropic

OpenAI

Zoom

Why Terms & Conditions Matter

AI meeting notetakers promise efficiency and free participants to focus on the discussion, but for anyone

handling confidential or proprietary data, a real risk lies in what happens to that data after capture.

Most vendors are built for product improvement and scale, so unless the terms say otherwise, your data

may be retained, reviewed by humans, or used to train models. For a life sciences company, this risk is

not theoretical. Capturing a confidential discussion is not, by itself, a public disclosure, but if that

information somehow becomes public, through a security breach or because the vendor’s terms permit

broad reuse, it can become prior art against your own later application and block your patent.

Trade secrets face the same threat. In addition to patents, life sciences companies derive much value

from trade secrets (e.g., assays, protocols, manufacturing know-how, etc.), but only if the company takes

reasonable measures to keep them secret. Routing that information through an AI tool whose terms allow

reuse or human access that is not carefully restricted can forfeit that trade secret protection. One court

recently dismissed a trade-secret claim on exactly that basis² .

Vendor marketing rarely addresses confidentiality in any meaningful detail and sometimes vendor

communications are inconsistent with their Terms & Conditions. So read the terms, privacy policy, and data processing addenda closely. These documents determine how your data is protected and whether the vendor acts purely as a processor. For most internal meetings, the core exposure is to your IP and trade secrets. But maintaining attorney-client privilege is important when a lawyer is involved, and the terms help decide whether privilege survives³ .

Terms to Read Closely

When reviewing the terms of an AI meeting notetaker, pay particular attention to clauses addressing:

Data ownership (who legally controls the content)

Retention and deletion (indefinite retention vs. user-controlled deletion)

AI training (opt-in vs. default training on your data)

Data encryption (in transit and at rest)

Human access (employees, contractors, quality review, etc.)

Disclosure obligations (subpoenas, government requests, corporate transactions)

Ambiguous or overly broad language in any of these areas can put your confidentiality and, where your

counsel is involved, your privilege at risk. Silence on these issues should be treated as a major risk,

particularly where unpublished inventions, trade secrets, or privileged communications are involved.

Bottom Line

Across many tools, AI is immensely helpful and its use is becoming almost unavoidable, but without

the proper legal protections it is a potential route to irreversible loss of patent rights and trade secrets and

waiver of privilege. A public disclosure of an invention, including one that inadvertently leaks from an AI

vendor’s systems, can destroy patentability and materially affect your company’s valuation. The Terms & Conditions should give you rock-solid confidence before you ever input confidential information into an

AI system.

Footnotes:

¹ When you use an AI notetaker, also confirm all participants have consented to recording, as many states

require the consent of all participants to record a conversation, and AI notetaker vendors have already

faced class-action claims on this basis. See, e.g., In re Otter.AI Privacy Litigation, No. 5:25-cv-06911

(N.D. Cal. 2025).

² See Trinidad v. OpenAI, Inc., No. 4:25-cv-06328 (N.D. Cal. Jan. 5, 2026) (dismissing DTSA claim

where the plaintiff disclosed material to a consumer AI platform under terms permitting reuse).

³ Lawyers who use these tools have their own professional-conduct obligations. See, e.g., ABA Standing

Comm. on Ethics & Pro. Resp., Formal Op. 512 (July 29, 2024). See also N.Y.C. Bar Ass’n Comm. on

Pro. Ethics, Formal Op. 2025-6 (Dec. 22, 2025). A non-lawyer’s use of a public AI tool can itself defeat

privilege. See United States v. Heppner, No. 1:25-cr-00503-JSR (S.D.N.Y. Feb. 17, 2026).

Disclaimer:

The content on this website (including all pages, articles, and comments) is not legal advice, and does not

and is not intended to form or constitute any attorney-client relationship. The content is not a solicitation

for business; it is for educational and entertainment purposes only, and reflects the personal views of the

author(s) only and not those of any past, present, or future client of DHL. Any content should be double-

checked for accuracy and current applicability, and liability is disclaimed for any error or omission.