On February 10, 2026, Judge Jed Rakoff of the Southern District of New York issued a ruling that should be required reading for every attorney practicing in Houston — whether you're handling energy litigation downtown, running a personal injury practice in The Heights, or managing complex commercial disputes that regularly touch the Southern District of Texas.
The ruling in United States v. Heppner found that 31 documents defendant Bradley Heppner created using Claude — Anthropic's AI assistant — were not protected by attorney-client privilege and were not covered by the work product doctrine. Judge Rakoff ordered them produced to the government.
This is widely believed to be the first ruling in which a federal court determined that a client's interactions with a publicly accessible AI tool — even when those interactions contained legally sensitive information — are not privileged communications. For Houston's legal community, where the volume of complex litigation across the Southern District of Texas is among the highest in the nation, this ruling has immediate, practical consequences.
What Happened in the Heppner Case
Bradley Heppner, facing federal charges, used Claude to process and analyze documents before sharing them with his legal team. He then turned those AI-generated outputs over to his attorneys, who incorporated them into the attorney-client relationship.
Heppner argued the documents should be protected. The government disagreed. Judge Rakoff sided with the government on three distinct grounds:
1. Claude is not a lawyer.
The AI platform cannot form a fiduciary attorney-client relationship. When a client communicates with Claude — no matter what they tell it, no matter how sensitive the information — no privilege attaches because the entity on the receiving end is not an attorney, is not supervised by an attorney, and owes no duty of confidentiality to the user as a matter of professional ethics or law.
2. No confidentiality exists.
Anthropic's privacy policy at the time notified users that it collects inputs and outputs to train its tools and may disclose data to third parties. Judge Rakoff held that this policy destroys any "reasonable expectation of privacy" required to claim privilege. If the AI company can use and share your data, the information isn't confidential in the legal sense.
3. No attorney direction.
Heppner used the AI on his own initiative, without being directed by counsel. The documents created did not reflect the mental processes or legal strategy of his attorneys at the time they were created. Client-created documents, standing alone, generally aren't privileged.
The Line Judge Rakoff Drew — And Why It Matters for Houston Attorneys
The ruling is a clear warning, but it also contains an important carve-out. Judge Rakoff noted:
"If counsel had directed Heppner to use the AI tool, 'Claude might arguably be said to have functioned in a manner akin to a highly trained professional who may act as a lawyer's agent within the protection of the attorney-client privilege.'"
Read that twice. The judge is drawing a direct line between consumer AI use by a client acting alone and attorney-directed AI use as part of legal representation. The former destroys privilege. The latter may preserve it.
This distinction is critical for how Houston attorneys counsel their clients and how they choose AI tools for their own practice. It's not about whether AI is used — it's about who directs it and within what structure.
Houston's Legal Landscape Makes This Ruling Especially Urgent
Houston is home to more than 15,000 licensed attorneys and one of the most active federal district courts in the country. The Southern District of Texas handles a massive docket of complex litigation — energy disputes, maritime claims, securities fraud, immigration cases, and major commercial litigation that regularly generates tens of thousands of discovery documents.
Smaller firms and solo practitioners across the greater Houston area — from Clear Lake to Katy, from Greenway Plaza to The Woodlands — are increasingly using AI tools to stay competitive. And their clients are doing the same.
Consider a scenario a Houston energy litigator might encounter today: a business owner facing a commercial dispute downloads a set of contracts, feeds them into ChatGPT or Claude, asks the AI to identify weaknesses in their position, then shares the output with their attorney. Under Heppner reasoning, those AI sessions — and potentially the output documents — are not privileged. If opposing counsel discovers the client used consumer AI and demands production, the attorney may have very little ground to stand on.
This is not a hypothetical. It is a live risk for every Houston attorney whose clients are sophisticated enough to use AI — and in 2026, that is most of them.
What Houston Solo Practitioners and Small Firms Should Do Right Now
Send a client advisory.
Draft a short memo for your active clients explaining that using consumer AI tools (ChatGPT, Claude, Gemini, Perplexity) to analyze case-related documents before or during representation can destroy privilege. Texas Rules of Professional Conduct Rule 1.1 now explicitly includes technological competence — proactive communication protects your clients and demonstrates it.
Review your own AI use.
If you are using consumer AI tools to analyze client documents, ask whether those tools meet the standard Heppner implies: attorney direction, confidentiality guarantees, and no data sharing for third-party training. Most consumer tools fail on at least one of these counts.
Understand what "attorney-directed AI" means practically.
Judge Rakoff's carve-out is not about which AI tool you use — it's about the structure of how AI is used in your practice. AI that operates under attorney supervision, within a platform designed for legal workflow, with verifiable data isolation and no training on your client data, is in a fundamentally different legal posture than a client typing questions into a chat interface.
How CaseIntel Is Built Differently
CaseIntel is an AI-powered legal discovery platform purpose-built for solo practitioners and small firms. The Heppner ruling's carve-out describes our architecture almost exactly.
Attorney Direction, Not Client Autonomy
CaseIntel is a tool attorneys use to direct document analysis — not a general-purpose chat interface. The attorney controls the platform, controls what documents are uploaded, and controls how the AI's output is used in the representation.
Verifiable Data Isolation
CaseIntel processes all documents through AWS Bedrock using Anthropic's Claude models within AWS's US-only infrastructure. Your client data is never used to train AI models, is never shared with third parties, and is isolated per-firm in a multi-tenant architecture.
Purpose-Built Privilege Detection
CaseIntel's automated privilege detection layer scans every uploaded document for attorney-client and work product indicators before analysis begins. Every flag is cited back to the specific language that triggered it.
Not a General Consumer Tool
Heppner's problem wasn't that he used AI — it was that he used a tool with no legal-workflow context, no attorney supervision, and no confidentiality architecture. CaseIntel is not that tool.
Practical Scenarios for Houston Attorneys
Personal Injury Practices
If your clients are using consumer AI to self-research their injuries, accident reports, or medical records before coming to you, those analyses may not be privileged. Build client intake language around this.
Energy and Commercial Litigation
Document volumes in Southern District of Texas cases regularly run to hundreds of thousands of pages. Using attorney-directed AI that maintains privilege protection and generates a defensible privilege log is no longer just efficiency — it's a professional responsibility issue.
Family Law and Estate Practices
Clients going through divorces often turn to AI to "figure out" what they're entitled to before retaining counsel. Advise clients early that those AI sessions may be discoverable.
Immigration Practices
In an environment of heightened enforcement in Houston, clients using AI to assess their situation before consultation are creating discoverable records. This is a critical client communication issue.
Courts Are Catching Up to AI
The Heppner ruling will not be the last of its kind. Expect similar motions in the Southern and Eastern Districts of Texas, State Bar of Texas guidance on attorney AI use, and malpractice exposure for attorneys who fail to advise clients about AI privilege risks.
Ready to use AI on the right side of the privilege line?
Start your free 14-day trial and see how attorney-directed AI discovery works — no enterprise complexity required.
Start Free TrialFrequently Asked Questions
What did the Heppner ruling say about AI and attorney-client privilege?
Judge Rakoff ruled that 31 documents created by a defendant using Claude AI were not protected by attorney-client privilege or work product doctrine. The court found that Claude is not a lawyer, consumer AI platforms lack confidentiality protections, and the documents were created without attorney direction.
Can Houston attorneys use AI tools while preserving attorney-client privilege?
Yes. Judge Rakoff explicitly noted that if counsel directs the use of an AI tool, the AI may function as the attorney's agent within the privilege. The key requirements are attorney direction, genuine confidentiality protections (no third-party data sharing), and documented attorney supervision.
What should Houston attorneys advise clients about using AI?
Attorneys should advise clients not to use consumer AI tools (ChatGPT, Claude, Gemini) to analyze case-related documents before consulting counsel. Such use may destroy privilege over those sessions and their outputs. This advisory should be added to engagement letters.
How does CaseIntel differ from consumer AI tools under the Heppner standard?
CaseIntel is an attorney-directed platform with no client-facing interface. All analysis is conducted at attorney direction through AWS Bedrock, with client data never used to train AI models or shared with third parties. This matches the privilege-preserving structure Judge Rakoff described.
This article is for informational purposes only and does not constitute legal advice. For guidance specific to your practice, consult your state bar's ethics counsel.
Further Reading
- United States v. Heppner, No. 1:23-cr-00251 (S.D.N.Y. Feb. 10, 2026) (Rakoff, J.)
- State Bar of Texas — Technology and the Practice of Law
- Houston Bar Association — Ethics & Professionalism Resources