# AI Confidentiality and Privilege: A Vendor Checklist Published and updated 2026-10-29 · RuleX editorial team · United States · American English Rule 1.6(c) requires reasonable efforts to prevent unauthorized disclosure of or access to client information, and ABA Opinion 512 asks lawyers to read an AI tool’s terms of use and privacy policy before entering client information. Answer twelve vendor questions in writing, obtain informed consent where a self-learning tool is involved and keep the record in the matter. RuleX publishes its terms and privacy policy for U.S. users. ## Rule 1.6(c) turns on reasonable efforts, so document the vendor review. Model Rule 1.6(c) requires a lawyer to make reasonable efforts to prevent the inadvertent or unauthorized disclosure of, or unauthorized access to, information relating to the representation of a client. State versions track the language; North Carolina’s 2024 Formal Ethics Opinion 1 and Florida’s Opinion 24-1 both build their AI analysis on their versions of the same duty. The standard is reasonable efforts, which means the lawyer’s diligence is the thing a regulator, a court or a client will ask to see. California’s 2026 guidance puts the diligence point directly: reasonable efforts require more than reliance on generalized marketing assurances, and they may include reviewing terms of use, privacy policies or vendor documentation and consulting qualified IT or cybersecurity professionals. It also observes that a user may agree to the terms of use and privacy policy simply by using the product, which is a reason to read them first. A written vendor review, dated and kept with the firm’s AI-use file, is the artifact that shows the efforts were made. ## Opinion 512 asks for informed consent for self-learning tools and a read of the vendor’s terms. ABA Formal Opinion 512 concludes that, because many self-learning generative AI tools are designed so that their output could lead directly or indirectly to the disclosure of information relating to a client’s representation, the client’s informed consent is required before that information is input. Consent is informed only if the lawyer explains why the tool is used, the extent of and specific information about the risk, the kinds of client information to be disclosed, the ways others might use it against the client’s interests and the tool’s benefits to the representation. Boilerplate added to an engagement letter is not enough. As a baseline, the opinion says all lawyers should read and understand the terms of use, privacy policy and related contractual terms of any generative AI tool they use, to learn who has access to the information that the lawyer inputs, or consult a colleague or outside expert who has analyzed them. Where the lawyer does not input information relating to the representation, as in general idea generation, consent under Rule 1.6 is not needed. State authorities add practical detail. Texas Opinion 705 lists four precautions: understand how the technology works, review and potentially renegotiate the terms of service, learn the tool’s data-security protections and train lawyers and staff. North Carolina advises lawyers to avoid inputting client-specific information into publicly available AI resources and notes that customized in-house tools derived from a public program may send inputs back to the central program for training, so the same vendor questions apply. ## Privilege is a separate question, and a federal court has ruled on a consumer AI tool. Confidentiality under the ethics rules and privilege under evidence law are different inquiries. North Carolina’s committee states that it does not opine on whether information shared with an AI tool affects the attorney-client privilege because that is a legal question outside the Rules of Professional Conduct, and it advises lawyers to research and resolve the privilege question before using a third-party AI program with client-specific information. Texas Opinion 705 adds that a lawyer’s prompts can expose the lawyer’s mental impressions to the tool. In United States v. Heppner, No. 1:25-cr-00503 (S.D.N.Y.), Judge Jed S. Rakoff ruled from the bench on February 10, 2026, and issued a written memorandum on February 17, 2026, that about thirty-one documents a criminal defendant created with a generative AI platform were protected by neither the attorney-client privilege nor the work product doctrine. The defendant had used the platform on his own, without counsel’s direction, to prepare reports on defense strategy after receiving a grand jury subpoena. The court reasoned that the exchanges were not communications with an attorney, were not confidential in light of the platform’s published privacy policy, which provided for collecting inputs and outputs, using them to train the model and disclosing data to third parties including governmental authorities, and were not made at counsel’s direction. It also noted that sharing the outputs with counsel afterward did not make them privileged. The memorandum draws a distinction that matters to practice: it stated that, had counsel directed the defendant to use the tool, the tool might arguably be said to have functioned like a highly trained professional acting as a lawyer’s agent. The takeaway is a workflow rather than a prediction. Use tools whose terms you have reviewed, use them within the representation at the lawyer’s direction and keep the record of that direction. The opinion is a district court ruling on specific facts; read it in full before relying on it in your forum. ## Twelve questions to answer from the vendor’s own terms. The questions below draw on Opinion 512’s list of vendor diligence points, the Texas and California precautions and the privacy-policy language that decided Heppner. Answer each from the vendor’s published terms or written confirmation, date the answers and keep them with the matter or the firm’s AI-use file. 1. Which terms govern my account: consumer, business or enterprise, and where are they published? 2. Does the provider use my prompts, uploaded documents or outputs to train or refine models, and does my plan change the answer? 3. Who can access my content: provider staff, sub-processors, other users or other lawyers in my firm behind an ethical wall? 4. How long does the provider keep my content, before and after I stop using the service, and does it assert rights in it? 5. How do I delete matter content, and what happens to the copies the provider keeps? 6. What security measures protect stored content, and will the provider notify me of a breach? 7. What does the provider do when it receives a subpoena or other service of process for my content, and will it notify me? 8. What does the privacy policy let the provider disclose to third parties or government authorities? 9. What liability terms does the provider set for a loss of my content? 10. Which AI providers process my content, what does each receive and where is it processed? 11. Can I separate content by matter and by user, and can I control who sees each matter? 12. Who approved the tool at my firm, and what training and information categories apply to it? ## What RuleX’s published privacy policy states about matter content. RuleX answers several of these questions in the section of its U.S. Privacy Policy titled “Legal content, AI, and confidentiality.” The policy states that content a user submits is processed to perform the requested function and, when the user retains it, to let the user consult their own work. It states that the AI providers RuleX uses are selected and configured under enterprise terms that exclude API inputs from model training, in accordance with the applicable provider terms, and that each AI provider receives only the content needed for the requested function. In the mobile app, RuleX asks for the user’s express authorization before content is first sent to third-party AI providers. The policy states that RuleX shares information with providers that help deliver the service, such as hosting, authentication, storage, email, payments, security and AI processing, solely for the corresponding purposes and subject to confidentiality and security obligations, and that internal access is limited to what is necessary to operate, protect or support the service. For retention, the policy states that history, documents, projects and transcripts are kept while the user or organization maintains them or until deletion is requested, subject to legal exceptions and backups, and that users can delete their history, documents, projects or transcripts with the available features or ask support for help. Read section 4 together with section 5 on providers and international transfers and section 6 on retention and deletion, then run the twelve questions against the text. The Terms of Use and Privacy Policy are published at rulex.org/united-states/terms and rulex.org/united-states/privacy, which gives you the reading Opinion 512 asks for before client information goes in. ## Record the review and give the client what they need to decide. Close the loop in the engagement and in the file. Describe the AI practices in the engagement agreement, obtain matter-specific informed consent when a self-learning tool is involved, and save the vendor review, the consent and the date. Opinion 512 identifies the engagement agreement as the logical place for the disclosure and for any client instruction on AI use. In RuleX, keep the vendor review, the client’s consent and the matter’s documents in the same case, with tasks and dates for the renewal of the review. Start the 7-day free trial, add a real matter and open the sources from the answer. ## Frequently asked questions ### What does Model Rule 1.6(c) require when I use an AI tool? It requires reasonable efforts to prevent inadvertent or unauthorized disclosure of, or unauthorized access to, information relating to the representation. Opinion 512 applies it to AI by asking you to weigh the sensitivity of the information, the likelihood of disclosure, the difficulty of safeguards and the effect of safeguards on the representation. ### When do I need the client’s informed consent to use an AI tool? Opinion 512 concludes that informed consent is required before information relating to a representation is input into a self-learning generative AI tool whose output could lead to disclosure of that information. Explain the reason for the tool, the specific risks, the information involved and the benefits; boilerplate in an engagement letter is insufficient. ### Does using an AI tool waive the attorney-client privilege? Privilege is a legal question that your forum’s law decides, and the ethics opinions leave it to the courts. In United States v. Heppner (S.D.N.Y. 2026), a defendant’s own use of a consumer AI platform without counsel’s direction was held neither privileged nor protected work product; the court noted that counsel-directed use might be analyzed differently. ### What should I read in an AI vendor’s terms? Read the terms of use, privacy policy and related contractual terms for model training on inputs, retention and deletion, access by staff and sub-processors, security and breach notice, handling of subpoenas, disclosure to third parties and liability. Opinion 512 asks lawyers to read these documents or consult someone who has. ### What does RuleX’s privacy policy say about AI providers? RuleX’s U.S. Privacy Policy states that its AI providers are selected and configured under enterprise terms that exclude API inputs from model training, in accordance with the applicable provider terms, and that each provider receives only the content needed for the requested function. Read the full policy at rulex.org/united-states/privacy. ## Sources - [ABA Formal Opinion 512 — Generative Artificial Intelligence Tools (July 29, 2024)](https://www.americanbar.org/content/dam/aba/administrative/professional_responsibility/ethics-opinions/aba-formal-opinion-512.pdf) - [State Bar of California — Practical Guidance for the Use of Generative AI in the Practice of Law (2026 edition)](https://www.calbar.ca.gov/Portals/0/documents/ethics/Generative-AI-Practical-Guidance.pdf) - [Texas Professional Ethics Committee — Opinion 705 (February 2025)](https://www.legalethicstexas.com/resources/opinions/opinion-705/) - [North Carolina State Bar — 2024 Formal Ethics Opinion 1 (adopted November 1, 2024)](https://www.ncbar.gov/for-lawyers/ethics/adopted-opinions/2024-formal-ethics-opinion-1/) - [S.D.N.Y. — United States v. Heppner, No. 1:25-cr-00503, Memorandum (Feb. 17, 2026), as reproduced by the Volokh Conspiracy](https://reason.com/volokh/2026/02/17/defendants-own-ai-legal-research-isnt-protected-by-attorney-client-privilege) - [RuleX — U.S. Privacy Policy](https://rulex.org/united-states/privacy) - [RuleX — U.S. Terms of Use](https://rulex.org/united-states/terms) ## Work on the matter in RuleX [Create a U.S. account](https://rulex.org/united-states?register=1) · [U.S. pricing](https://rulex.org/united-states/pricing) ## Representation This Markdown represents the same public page. Browsers receive HTML at its canonical URL. - **Canonical (HTML):** https://rulex.org/united-states/guides/ai-tool-confidentiality-vendor-checklist - **This representation:** https://rulex.org/united-states/guides/ai-tool-confidentiality-vendor-checklist.md - **Home:** https://rulex.org/united-states - **llms.txt:** https://rulex.org/llms.txt