For those facing federal or state investigations, the “right to remain silent” has entered a complicated new era. While the integration of Artificial Intelligence into daily life has become seamless, a landmark 2026 ruling has turned these digital tools into a potential liability for the accused.

In United States v. Heppner, the Southern District of New York addressed a “question of first impression” that is already sending ripples through the legal community: Is your private conversation with an AI chatbot protected by attorney-client privilege?

The court’s answer was a resounding no. For defendants, this decision means that a casual query to a chatbot could become the prosecution’s strongest piece of evidence—effectively making the AI a “witness” for the government.

The Case: Bradley Heppner’s “Digital Roadmap”

The case involved Bradley Heppner, a financial executive who used Claude, a generative AI platform, to research the government’s potential case and outline his defenses before formally speaking with his legal team.

When federal agents executed a search warrant, they seized his devices and discovered his AI-generated research. Despite claims from his defense team that these conversations were privileged “work product,” the court granted the government’s motion to compel production.

Why the Court Rejected the Privilege Claim

Judge Jed Rakoff identified three primary reasons why these AI prompts failed to meet the standards of confidentiality:

1. AI is Not an Attorney: Under the law, attorney-client privilege requires communication with a licensed human professional. An AI, no matter how sophisticated, cannot be a “confidant” in the eyes of the judiciary.
2. The Privacy Policy Waiver: By using consumer AI tools, users agree to Terms of Service that explicitly permit the company to store and share data with authorities. This destroys the “reasonable expectation of privacy” required for legal protection.
3. Lack of Attorney Agency: Because the AI was not acting as an “agent” of a law firm (like a paralegal or expert witness might), the interaction remained a strictly commercial, third-party transaction.

Does AI usage waive my Fifth Amendment rights?

The Fifth Amendment protects you from being compelled to be a witness against yourself. However, the Heppner ruling clarifies a devastating distinction: the Fifth Amendment protects your mind, but not necessarily your cloud storage. When you type a prompt into an AI, you are creating a record. Once that record exists on a third-party server, the government can subpoena that data without “compelling” you to speak.

The “act of production” doctrine might protect you from having to hand over a password, but it rarely protects the data itself once the government obtains it from the service provider. By talking to an AI, you are essentially whispering your secrets to a digital stenographer who is legally obligated to hand their notes to the FBI upon request.

How does the Third-Party Doctrine affect AI chats?

The “Third-Party Doctrine” is a cornerstone of American privacy law that states people who voluntarily give information to third parties—such as banks, phone companies, or internet service providers—forfeit any “reasonable expectation of privacy” in that information.

In United States v. Heppner, the court solidified the application of this doctrine to Large Language Models (LLMs). Because the user is sending data to a commercial entity to be processed, the Fourth Amendment’s protection against unreasonable searches is significantly weakened. The moment your packet of data hits a remote server, it loses its “private” status.

What is the difference between “Consumer AI” and “Legal AI”?

A common point of confusion is whether any use of AI in a legal case is dangerous. The answer lies in the architecture of the software.

• Consumer AI (ChatGPT, Claude, Gemini): These are built for the general public. Their business models often involve using your data to train their models, and their Terms of Service (ToS) are written to protect the company, not your legal rights.
• Enterprise/Legal AI: These are closed environments. When a law firm uses an Enterprise version of these tools, they often sign “Zero-Retention” agreements. In these cases, the AI can be considered an “agent of the attorney,” similar to a private investigator or a forensic accountant.

The Heppner ruling specifically targeted a defendant who used a consumer-grade tool on his own initiative. Had his attorney directed the research through a secure, firm-managed portal, the outcome might have been different.

Sharing is Not Shielding

A critical takeaway for defendants is the rejection of “retroactive privilege.” Heppner’s defense argued that because he eventually shared the AI logs with his lawyers, they should be protected.

The court dismissed this, stating that sharing non-privileged material with a lawyer does not change it into privileged communication. If a document is discoverable in your hands, it remains discoverable even after you give it to your attorney. This is known as the “pre-existing document” rule. You cannot hide digital evidence by handing it to a lawyer.

Can the SBI or FBI subpoena my AI history?

Yes. The State Bureau of Investigation and federal agencies like the FBI have become increasingly sophisticated in digital forensics. If they have probable cause to believe a crime was committed, a search warrant for your Google or OpenAI account is standard procedure.

State prosecutors are already using search history and GPS data in high-profile felony cases. The Heppner case provides the blueprint for them to now include “Prompt History” as a primary source of evidence. Imagine a prosecutor reading your own words—where you asked an AI “how to hide assets” or “what are the penalties for wire fraud”—to a jury. Under North Carolina law, the state has broad authority to issue subpoenas for electronic service providers to obtain such records pursuant to N.C.G.S. § 15A-298.

Is my AI chat history considered “Hearsay” in court?

A major hurdle for any evidence is whether it is “hearsay”—an out-of-court statement offered to prove the truth of the matter asserted. However, your own prompts to an AI are classified as “Statements by a Party-Opponent” under the Rules of Evidence. In North Carolina, this falls under the hearsay exception found in N.C.G.S. § 8C-1, Rule 801(d), meaning they are fully admissible.

Furthermore, the AI’s responses may be admitted to provide context for your questions or as “non-hearsay” to show your state of mind. In the Heppner case, the executive’s detailed questions about specific statutes showed “consciousness of guilt,” a powerful tool for any prosecutor.

The “Heppner Trap” for High-Tech Professionals

Professionals in North Carolina’s technology and banking hubs are often early adopters of emerging tools. However, a “Heppner trap” can inadvertently ensnare white-collar defendants who are accustomed to solving complex problems through technology. While using AI to organize thoughts or research statutes may seem efficient, doing so on a consumer platform may create a digital trail that lacks legal protection.

Moreover, the legal standards regarding “work product” are exacting. In North Carolina, the work-product doctrine—as outlined in N.C.G.S. § 1A-1, Rule 26(b)(3)—typically requires that materials be prepared “in anticipation of litigation” by or for a party’s representative. Most jurisdictions generally require that such research be conducted at the specific request of counsel to qualify for protection.

If a defendant begins independent legal research via AI before retaining an attorney, they may unintentionally generate discoverable data. Under the current legal framework, information created outside the direction of counsel is often viewed as a voluntary disclosure to a third-party service provider, which can then be subpoenaed by the prosecution.

What is “Prompt Engineering” as evidence?

In the Heppner trial, the prosecution didn’t just look at the answers the AI gave; they looked at how Heppner engineered his prompts. By analyzing his step-by-step questioning, the government was able to show a “pattern of intent.”

For example, if a user starts with a broad question like “What is the law on taxes?” and narrows it down to “How to explain $50,000 in offshore transfers to the IRS,” the progression itself becomes evidence of a specific criminal intent. The digital trail of a prompt history is often more revealing than a final document because it shows the evolution of a person’s thoughts.

Strategic Information: Protecting Your Rights in the Digital Age

The Heppner ruling serves as a vital reminder that digital convenience can inadvertently compromise a legal defense. To help safeguard constitutional protections, consider the following principles:

• Avoid Independent AI Research: Using public chatbots to “brainstorm” defenses or summarize evidence can create a discoverable digital roadmap for the prosecution.
• Prioritize Secure Environments: If a legal team utilizes AI, it should ideally be within a secure, Enterprise-grade environment that offers “agent-of-counsel” status. These tools are designed to maintain the privacy necessary for the work-product doctrine.
• Consult Your Attorney on Past Use: If you have already utilized AI tools to research your situation, notify your legal counsel immediately. An attorney can assess potential “waiver” issues and determine the best path forward before the discovery process begins.
• Maintain Digital Mindfulness: In a connected era, it is safest to operate under the assumption that data transmitted to third-party servers may be subject to future legal recovery. Protecting your defense starts with a strategy of digital discretion.

Can I delete my AI history to protect myself?

Proceed with extreme caution. Once you are aware of a potential investigation, deleting your AI history could be considered obstruction of justice or spoliation of evidence. Destroying evidence can lead to separate criminal charges and a “jury instruction” that allows the jury to assume the deleted evidence was harmful to you. In North Carolina, the intentional destruction of evidence can also lead to criminal liability for Altering, Destroying, or Stealing Evidence as defined in N.C.G.S. § 14-221.1.

Instead of deleting, you must stop using the platforms and consult with a lawyer who can handle the “preservation” of that data in a way that minimizes its impact.

The Future: Will the Law Evolve?

There is currently no legislation that would extend attorney-client privilege to AI. While some privacy advocates argue that the law needs to catch up to technology, the current legal climate favors the government’s ability to access third-party data. Until a statute is passed or a supreme court rules otherwise, the Heppner precedent remains the most significant warning for residents.

Why “Incognito Mode” doesn’t help you

Many users mistakenly believe that using “incognito” or “private” browsing modes, or “temporary chats” in AI interfaces, protects them. This is a dangerous misconception. These modes primarily prevent data from being saved on your local device, but the data is still transmitted to and stored on the company’s servers. A federal subpoena goes to the company, and they will provide the logs regardless of whether you used a “private” tab. The North Carolina Electronic Surveillance Act, found in N.C.G.S. § 15A-286 et seq., provides the framework under which state law enforcement can access stored electronic communications.

Summary of the “Heppner Ruling” Impact

Feature	Consumer AI (Post-Heppner)	Attorney-Led AI
Privilege Status	None (Waived)	Protected (Work Product)
Searchable by Authorities?	Yes, via Subpoena/Warrant	No (Highly Protected)
Admissible in Court?	Yes, as Party Statement	No
Privacy Expectation	Low (Commercial ToS)	High (Legal Ethics)

Frequently Asked Questions: AI and Your Defense

1. Is my conversation with an AI chatbot protected by attorney-client privilege? 

No. Following the United States v. Heppner ruling, courts have determined that attorney-client privilege only applies to communications with licensed human professionals. Because AI is a third-party commercial service and not a legal professional, your interactions are not protected.

2. Can the FBI or state police subpoena my AI chat history? 

Yes. Federal and state agencies can issue subpoenas or search warrants to AI service providers like OpenAI or Google. Under the Third-Party Doctrine, once you share information with these companies, you lose a “reasonable expectation of privacy,” making that data accessible to law enforcement.

3. Does using a “Temporary Chat” or “Incognito Mode” protect my data from investigators? 

No. These features generally prevent data from being saved to your local device or browser history, but the information is still transmitted to and stored on the company’s servers. Law enforcement subpoenas are directed at the service provider, who can still produce the logs of those sessions.

4. Can my AI prompts be used as evidence of my intent in court? 

Yes. Your specific questions and “prompt engineering” can be used to show a “pattern of intent.” For example, narrowing a search from general law to specific methods of bypassing regulations can be presented to a jury as evidence of “consciousness of guilt.”

5. How does the Fifth Amendment “right to remain silent” apply to AI records? 

While the Fifth Amendment protects you from being forced to speak, it does not necessarily protect records already stored on a third-party server. Because you voluntarily typed the information into the AI, the government can obtain those existing records from the provider without “compelling” you to testify against yourself.

6. What is the difference between “Consumer AI” and “Legal AI” regarding privacy? 

Consumer AI (like the free or standard versions of ChatGPT or Claude) operates under terms of service that allow for data storage and sharing. Enterprise or “Legal AI” used by law firms often includes “Zero-Retention” agreements, where the AI acts as an “agent of the attorney,” potentially qualifying the data for protection under the work-product doctrine.

7. If I share my AI chat logs with my lawyer, do they then become privileged? 

No. Under the “pre-existing document” rule, if a document was not privileged when it was created, you cannot make it privileged simply by handing it to an attorney. If the prosecution could have seized the logs from you or the AI company, they can still seize them from your lawyer.

8. Are my AI chats considered “hearsay” and therefore inadmissible in court? 

Generally, no. Your prompts are considered “Statements by a Party-Opponent,” which is a specific exception to the hearsay rule. This means your own words can be admitted as direct evidence against you in both federal and North Carolina state courts.

9. Can I delete my AI history if I think I am under investigation? 

You should consult an attorney before deleting any data. If you are aware of a potential investigation, deleting records can be classified as obstruction of justice or “spoliation of evidence,” which can lead to additional criminal charges or a jury instruction to assume the deleted data was incriminating.

10. How can I research my legal situation without creating a “digital roadmap” for prosecutors? 

You should cease all personal AI research regarding your case immediately. Instead, conduct all inquiries through your legal counsel. Your attorney can use secure, firm-sanctioned tools that ensure your research is conducted under the protection of the work-product doctrine and attorney-client privilege.

Strategic Defense in a Digital Age

Navigating the intersection of emerging technology and constitutional law requires a defender who understands the technical mechanics behind the software. Attorney Patrick Roberts provides a background that is uniquely suited to these modern challenges.

Professional Background and Credentials

• Former Wake County Prosecutor: Before entering private practice, Patrick Roberts served the community as a prosecutor in Wake County, gaining an internal perspective on how the state builds its cases.
• Engineering Foundation: An alumnus of Johns Hopkins University, Patrick Roberts earned his degree in Engineering, providing him with the technical analytical framework to deconstruct how AI models process and store data.
• Legal Training: He earned his Juris Doctor (J.D.) from Duke University School of Law, which is currently ranked Number 9 in Criminal Law for the 2026-2027 period.
• Elite Trial Training: He is a graduate of the National Criminal Defense College (NCDC) and was trained at the Gerry Spence Trial Lawyers College, institutions dedicated to the highest levels of courtroom advocacy.
• Professional Affiliations: He is a Lifetime Member of the National Association of Criminal Defense Lawyers (NACDL).
• Technical Commitment: Patrick Roberts is continuously honing his skills and updating his knowledge of technological breakthroughs in the criminal defense space to ensure his clients receive the most current protections available. In 2025, he attended the NCDC Cross-Examination Intensive training.

Protecting Constitutional Rights 

Patrick Roberts uses his technical engineering degree, rigorous legal education, and extensive trial training to provide a technical shield for his clients. He helps protect your rights by:

• Strategic Consultation: Advising clients on how to avoid inadvertently waiving their Fifth Amendment rights or attorney-client privilege when using digital tools.
• Technical Evidence Analysis: Scrutinizing the methods used by law enforcement to obtain digital data and challenging the validity of subpoenas targeting AI logs.
• Implementing Secure Protocols: By using secure, firm-sanctioned environments, Roberts ensures that research conducted under his direction is shielded by the Work-Product Doctrine, moving your data from “discoverable” to “privileged.”

---

What Others Are Saying

Peer Review

“Patrick Roberts is undoubtedly one of the best attorney advocates in the Triangle area. His experience as a former prosecutor is invaluable to any client seeking an attorney who knows the ropes. His honesty, integrity, and straightforwardness are qualities every client should seek in an attorney. I would, without hesitation, strongly recommend Mr. Roberts to anyone seeking an outstanding criminal defense lawyer.” — Verified Peer Endorsement via Avvo

Client Review

“Mr. Roberts is kind, professional, and exceptionally competent. His dedication to restoring normalcy to our daughter’s life was evident in everything he did, and our family is deeply grateful for his guidance and advocacy. We would highly recommend Attorney Patrick Roberts to anyone in need of skilled, trustworthy, and compassionate legal representation. He is exactly the type of attorney you want standing beside you when everything feels uncertain.” — Verified Client Review via Avvo

Disclaimer: Testimonials and peer reviews are for informational purposes only and do not guarantee or predict the outcome of your legal matter. Every case is unique and must be evaluated on its own merits. All endorsements featured on this site are actual comments from clients and peers.

---

The New Rule of Silence

The adage “anything you say can and will be used against you” now extends to your keyboard. United States v. Heppner proves that the government doesn’t need a wiretap when a defendant provides a written roadmap of their defense to a commercial server.

The “Right to Remain Silent” must now include the “Right to Remain Offline” regarding legal matters. In the age of AI, the most effective defense strategy starts with digital silence. If you feel the urge to “just ask” a chatbot about your legal troubles, remember that your “digital assistant” may very well be the star witness for the prosecution.

In this high-tech legal landscape, you need a lawyer who speaks the language of the code as fluently as the language of the law. Patrick Roberts Law  combines the technical foundation of a Johns Hopkins engineer with the training of a Duke Law graduate and the experience of a former Wake County prosecutor to defend your case.

If you are facing criminal charges in North Carolina, contact Patrick Roberts Law for a confidential consultation. All inquiries are handled with the highest level of professional discretion and strict confidentiality.

*To maintain the high standard of advocacy required for high-stakes digital and constitutional defense, our firm accepts a limited number of cases. This ensures that every client receives the rigorous attention and deep technical analysis their situation demands. 

---

Disclaimer: The information on this website is for general informational purposes only. Nothing herein should be taken as legal advice for any individual case or situation. Contacting us via this website, email, or contact form does not create an attorney-client relationship. Case results depend on a variety of factors unique to each case; prior results do not guarantee a similar outcome.

Tags: Artificial Intelligence, Digital Evidence
Posted in Artificial Intelligence | No Comments »