motion

Read

would

<p>Consulting an AI chatbot for legal matters can see your chat history used against you by opposition lawyers… and potentially risks waiving attorney-client privilege with your own lawyer.</p><p>AI tools are affordable, if unreliable, sources of legal advice and their use raises uncomfortable questions about privilege, discoverability and how much <a href="https://cointelegraph.com/magazine/europe-chat-control-germany-oppose-scan-messages/" rel="noopener noreferrer nofollow">control users really have over their data</a>.</p><p>Those risks are already materializing in courts. AI companies have been compelled to retain user chat logs that would otherwise have been deleted.</p><figure class="wp-block-image aligncenter size-full is-resized"><img width="797" height="937" src="https://s3.magazine.cointelegraph.com/magazine/wp-content/uploads/2026/01/Lawyer-fined-AI-citations.jpg" alt="Lawyer fined AI citations" class="wp-image-56259"><figcaption class="wp-element-caption"><em>AI use in the legal field is increasing, but it does not always result in greater efficiency. (</em><a href="https://x.com/rryssf_/status/2013161394162778386" rel="noopener noreferrer nofollow"><em>Robert Youssef</em></a><em>)</em></figcaption></figure><p>Magazine brought those questions to Charlyn Ho, CEO of law and consulting firm Rikka, to understand how existing legal frameworks apply to AI tools and where the risks lie.</p><p>This conversation has been edited for clarity and length.</p><h2 class="wp-block-heading" id="magazine-does-using-ai-risk-waiving-attorney-client-privilege">Magazine: Does using AI risk waiving attorney-client privilege?</h2><p><strong>Ho: </strong>The privilege issue is complicated and not fully defined yet, but in general, attorney-client privilege is waived if you voluntarily disclose privileged information to a third party that is unrelated to the matter.</p><figure class="wp-block-image aligncenter size-full is-resized"><img width="805" height="422" src="https://s3.magazine.cointelegraph.com/magazine/wp-content/uploads/2026/01/OpenAI-Terms-of-services.jpg" alt="Open AI terms and conditions explains how it may use data" class="wp-image-56265"><figcaption class="wp-element-caption"><em>OpenAI can use customer data to train its AI models unless users opt out. (</em><a href="https://openai.com/policies/how-your-data-is-used-to-improve-model-performance/" rel="noopener noreferrer nofollow"><em>OpenAI</em></a><em>)</em></figcaption></figure><p>The default for public AI tools is that the model is training on your data. If you go to the terms and conditions, it will say that the model will train on the data. There is a very possible chance of voluntary disclosure of otherwise privileged information to a third party. This does change when you upgrade to enterprise models.</p><h2 class="wp-block-heading" id="magazine-where-does-the-line-get-drawn-between-tools-that-preserve-privilege-and-tools-that-risk-waiving-it">Magazine: Where does the line get drawn between tools that preserve privilege and tools that risk waiving it?</h2><figure class="wp-block-image alignleft size-full is-resized"><img width="902" height="565" src="https://s3.magazine.cointelegraph.com/magazine/wp-content/uploads/2026/01/AI-chatbot-court.jpg" alt="lawyer on X explains that privileges don't extend to AI chat bots. " class="wp-image-56250"><figcaption class="wp-element-caption"><em>An AI privilege doesnt exist yet. (</em><a href="https://x.com/JoeCarlasare/status/2005798129849806906" rel="noopener noreferrer nofollow"><em>Joe Carlasare</em></a><em>)</em></figcaption></figure><p><strong>Ho:</strong> This is where it becomes fact- and circumstance-based. With Microsoft Word on a CD, you download the program to your computer and draft your motion locally. That remains privileged because it is protected within your own environment and is not disclosed to a third party.</p><p>Fast forward to Microsoft 365. The document is now in the cloud, but Microsoft says that its customer data. The customer can put certain protections around it. There are obviously certain exceptions.</p><p>With AI, that <a href="https://cointelegraph.com/magazine/no-civil-protection-crypto-china-300k-list-coins-hong-kong-asia-express/" rel="noopener noreferrer nofollow">contractual protection is no longer automatic</a>. Customer data is your data. You can set parameters, delete it and control how it is used. It does not persist in memory indefinitely.</p><p>Im writing a book on how to use AI in a way that gives users greater confidence that privilege is not waived. A significant part of it focuses on ensuring that contracts with AI providers clearly state that the data belongs to the user, that the provider is not training on that data and that appropriate security measures are in place. These include encryption and limited retention periods.</p><div class="article-suggest"><div class="article-suggest__title display3"> Read also </div><div class="article-suggest__items"><div class="article-suggest__item"><a href="https://cointelegraph.com/magazine/devil-takes-bitcoin-book-review-jake-adelstein/" class="article-suggest__subtitle display4" rel="noopener noreferrer nofollow"><span>Features</span><p>Review: The Devil Takes Bitcoin, a wild history of Mt. Gox and Silk Road</p></a></div><div class="article-suggest__item"><a href="https://cointelegraph.com/magazine/bitcoin-kyle-chasse-venture-capitalist-youtube-strike-cointelegraph/" class="article-suggest__subtitle display4" rel="noopener noreferrer nofollow"><span>Features</span><p>Bitcoin OG Kyle Chass is one strike away from a YouTube permaban</p></a></div></div></div><h2 class="wp-block-heading" id="magazine-if-a-client-uses-chatgpt-to-help-with-a-legal-dispute-or-to-draft-internal-notes-could-that-material-be-discoverable-by-courts-or-opposing-counsel">Magazine: If a client uses ChatGPT to help with a legal dispute or to draft internal notes, could that material be discoverable by courts or opposing counsel?</h2><p><strong>Ho: </strong>Yes, definitely.</p><p>There was the OpenAI copyright litigation. The court ordered OpenAI to preserve and segregate all output log data that would otherwise be deleted, overriding user deletion requests.</p><figure class="wp-block-image aligncenter size-full"><img width="811" height="290" src="https://s3.magazine.cointelegraph.com/magazine/wp-content/uploads/2026/01/OpenAI-ordered-to-produce-20-million-de-identified-consumer-chat-logs.jpg" alt="Court orders open AI to produce 20 million de-identified consumer chat logs" class="wp-image-56262"><figcaption class="wp-element-caption"><em>OpenAI was recently ordered to disclose 20 million ChatGPT logs. (</em><a href="https://storage.courtlistener.com/recap/gov.uscourts.nysd.640396/gov.uscourts.nysd.640396.734.0.pdf" rel="noopener noreferrer nofollow"><em>In re OpenAI, Inc. Copyright Infringement Litigation/CourtListener</em></a><em>)</em></figcaption></figure><p>Lets say, hypothetically, you type into ChatGPT, How do I get away with this fraudulent thing, and then you think, Oh crap, thats not good. I want to delete that. The court could override the users deletion request and compel OpenAI to preserve the records.</p><p>These are evolving issues, but AI-related inputs, outputs or prompts would be treated the same way that Microsoft, Google or Amazon could be sent a subpoena to preserve records.</p><div class="article-suggest"><div class="article-suggest__title display3"> Read also </div><div class="article-suggest__items"><div class="article-suggest__item"><a href="https://cointelegraph.com/magazine/tokenizing-music-royalties-nfts-taylor-swift-3lau-nas-the-weeknd/" class="article-suggest__subtitle display4" rel="noopener noreferrer nofollow"><span>Features</span><p>Tokenizing music royalties as NFTs could help the next Taylor Swift</p></a></div><div class="article-suggest__item"><a href="https://cointelegraph.com/magazine/hundreds-millions-spent-marketing-crypto-sports-fans-wasted/" class="article-suggest__subtitle display4" rel="noopener noreferrer nofollow"><span>Features</span><p>Billions are spent marketing crypto to sports fans Is it worth it?</p></a></div></div></div><h2 class="wp-block-heading" id="magazine-do-you-foresee-the-development-of-online-ai-lawyers-that-are-protected-from-this-discoverability">Magazine: Do you foresee the development of online AI lawyers that are protected from this discoverability?</h2><figure class="wp-block-image alignright size-large is-resized"><img width="487" height="1024" src="https://s3.magazine.cointelegraph.com/magazine/wp-content/uploads/2026/01/DoNotPay-first-robot-lawyer-487x1024.jpg" alt="Screen showing DoNotPay services on mobile app" class="wp-image-56256"><figcaption class="wp-element-caption"><em>Despite its legal issues, DoNotPay reported a $210-million valuation in 2021. (DoNotPay)</em><br></figcaption></figure><p><strong>Ho: </strong>There were already so-called AI lawyers even before the generative AI explosion. DoNotPay, for example, helped people contest traffic violations on their own without a lawyer, and it grew quite large. But it also faced <a href="https://storage.courtlistener.com/recap/gov.uscourts.cand.410868/gov.uscourts.cand.410868.1.0.pdf" rel="noopener noreferrer nofollow">challenges</a> related to the unauthorized practice of law.</p><p>In the US, there are protectionist measures that shield the legal profession from outsiders. You cannot practice law without a license, and that includes AI. As things stand, a human lawyer with a valid license must be involved for something to be considered legal advice.</p><p>Until the bar changes the rules, I do not foresee a truly autonomous AI lawyer. There may be agents that help people navigate legal issues, but they will not be deemed to be providing legal advice.</p><h2 class="wp-block-heading" id="magazine-in-crypto-enforcement-cases-where-blockchain-transactions-are-public-but-intent-is-unclear-could-regulators-or-courts-rely-on-ai-chat-records-to-assess-intent-or-knowledge">Magazine: In crypto enforcement cases, where blockchain transactions are public but intent is unclear, could regulators or courts rely on AI chat records to assess intent or knowledge?</h2><p><strong>Ho:</strong> In a hypothetical future, I could see AI being used to help infer intent, but I do not think AI is the only way to do so.</p><p>With illicit activity, you can see the transaction on the blockchain, but to pursue an enforcement action, intent to commit fraud, to defraud or to engage in <a href="https://cointelegraph.com/magazine/cryptos-impossible-choice-when-privacy-and-aml-laws-conflict/" rel="noopener noreferrer nofollow">money laundering has to be derived</a> from something else.</p><p>AI chats would be just one element of evidence. I do not see them as fundamentally different from other forms of evidence, except that they may be more powerful.</p><p>There is often a tendency to treat new technologies as fundamentally different or as breaking existing models. From a legal perspective, I see it differently. The law does not move that quickly. New technologies are typically assessed within existing legal frameworks.</p><p>AI transaction logs, to the extent that they constitute evidence in criminal prosecutions or other enforcement actions, are not materially different from existing methods used to <a href="https://cointelegraph.com/magazine/max-burwick-lawyer-sue-pump-fun-ambulance-chaser-of-crypto/" rel="noopener noreferrer nofollow">collect evidence in blockchain-related cases</a>.</p><div class="subscribe subscribe--inner"><div class="container"><div class="subscribe__inner"><div class="subscribe__content"><div class="subscribe__title display2"> Subscribe </div><div class="subscribe__desc text-l"> The most engaging reads in blockchain. Delivered once a week. </div><div class="wpcf7 no-js" id="wpcf7-f13672-o21"><div class="screen-reader-response"><ul></ul></div><p>Email address</p><p>SUBSCRIBE</p><div class="wpcf7-response-output"></div></div></div><div class="subscribe__img"><img src="https://images.cointelegraph.com/cdn-cgi/image/format=auto,onerror=redirect,quality=80,width=373,dpr=1,gravity=0.5x0.5/https://s3.magazine.cointelegraph.com/magazine/wp-content/uploads/2022/10/reading-copy.png" alt="Subscribe to Magazine by Cointelegraph Newsletter."></div></div></div></div>

The critical reason you should never ask ChatGPT for legal advice

Table of Contents

Cointelegraph

Other articles published on Jan 22, 2026

Vitalik Buterin Says Decentralized Social Is How Ethereum Outlives Its Creators

Can HBAR Hit $1? What It Would Take

JasmyChain Completes Mainnet Migration And Enters Production

What to Expect from Bee Network in 2026

Kaisar Network Explained: Earn From Your Idle Hardware

BNB Chain Launches $650k Trenching Season Across Five Trading Platforms

Join our free newsletter for daily crypto updates!