Consumer artificial intelligence (AI) platforms such as ChatGPT, Claude, and Gemini are rapidly growing in popularity. People turn to them with questions of every kind, from simple to complex, including questions about legal issues. AI can be a powerful tool, but it also comes with significant risks.

In this article, we outline the most common issues we see when people use AI for legal questions involving patents, trademarks, and copyrights. Consider these issues before you rely on AI for legal advice, and especially before you ask AI to weigh in on advice from your attorney.

Key Takeaways

• AI can be wrong, and it is difficult to tell how an AI tool arrived at its answer.
• Consumer AI platforms generally do not promise confidentiality for what you share with them.
• Sharing unfiled inventions, business ideas, or trademarks with AI may put their confidentiality at risk.
• Sharing your attorney’s advice with AI risks waiving the attorney-client privilege.
• California Senate Bill 574, effective January 1, 2027, restricts how California-licensed attorneys may use generative AI.

AI Is a Powerful Tool for Answering Questions

No one debates that AI can answer many complex questions. In many ways, AI is a supercharged search engine. It ingests much of the content on the internet and, when prompted with a question, synthesizes that content into a single, plain-language answer that a regular person can understand. Instead of page after page of search results, some relevant and some less so, AI provides one answer. Some AI tools also cite their sources, which can be helpful.

But relying on AI models can be risky: the answer may be incorrect, and your privacy/confidentiality/privilege may not be protected.

Risk #1: AI Answers Can Be Wrong

AI is a statistical model that tries to determine the correct answer based on a number of inputs. Sometimes it picks the wrong answer.

Of course, human beings give incorrect answers too, all the time. But over the millennia, people have developed better tools for judging the quality of a human answer, such as the reputation of the person answering, their educational credentials, their professional licenses, and their years of experience in the field. Judging the quality of an AI answer is much harder, because it is difficult to identify the inputs and logic the model used to arrive at that specific result.

In some cases, AI helps clients identify and frame the key questions about their legal issue, which leads to a more productive and efficient consultation. In other cases, we have seen AI provide incorrect answers that create more questions, require time spent running down incorrect information, cause disputes between lawyer and client, and generally waste energy. AI can be a source of information, but that information should be treated with healthy skepticism.

Risk #2: Loss of Confidentiality and Attorney-Client Privilege

Confidentiality is a core tenet of the attorney-client relationship, and it applies to prospective, current, and past clients. As attorneys, we must keep the communications our clients provide to us confidential, and we may disclose client confidences only in very limited circumstances. The attorney-client privilege also protects confidential communications between attorney and client. Generally, neither the attorney nor the client can be forced to testify about privileged communications or to produce them in litigation. Together, confidentiality and the attorney-client privilege allow clients to be honest with their attorneys and seek legal advice without fear that their communications will later be disclosed to their detriment.

Consumer AI models, by contrast, generally provide no agreement to keep their users’ content confidential. A quick review of the OpenAI Terms of Use and Privacy Policy reveals no agreement or duty to keep communications confidential comparable to the duty an attorney is ethically obligated to provide a client. In fact, OpenAI clearly states that it can use your inputs and outputs (your content) to “provide, maintain, develop, and improve” its services, which means to train its AI. That does not mean it will keep your information confidential. OpenAI’s Privacy Policy also gives it broad rights to disclose your personal information to a number of third parties for various business purposes. Its terms even state that users should not use output for “any purpose that could have a legal or material impact on that person, such as making credit, educational, employment, housing, insurance, legal, medical, or other important decisions about them.”

Most intellectual property clients come to us with new business ideas, new business names, inventions, or other ideas that they want to keep confidential until they protect these assets and launch their business. Disclosing those ideas to an AI model that can use the information for any number of purposes may put the confidentiality of the idea at risk. In addition, disclosing attorney-client privileged communications to third parties not covered by that privilege can be considered a waiver of the privilege.

So what does this mean for you? General legal questions to AI, such as “What’s the process for a trademark application?” are generally acceptable, as long as you remember that AI answers can be incorrect. Disclosing confidential business ideas, inventions, trademarks before filing, or any other confidential information to an AI model risks the confidentiality of that information. And disclosing advice you have received from an attorney runs the risk of waiving the attorney-client privilege, which could become a significant issue down the road.

California SB 574: New Rules for Attorneys Using Generative AI

California Senate Bill 574, which regulates the use of generative AI by attorneys, goes into effect January 1, 2027. The law prohibits California-licensed attorneys from:

• Delegating the practice of law to generative AI. Legal advice must be based on the attorney’s knowledge of the law, not just an answer provided by AI;
• Providing confidential, personally identifying, or non-public information to AI; and
• Filing briefs in court with unverified citations provided by AI.

Attorneys must also take reasonable steps to verify the accuracy of generative AI outputs. The law contains a few other provisions, but these are the ones most relevant to this article.

In short, attorneys cannot delegate their professional judgment to AI, provide confidential information to AI, or use AI-generated legal citations without verification. If you are seeking quality advice on a legal question, it is worth asking why you would take any of these prohibited actions yourself.

The Bottom Line

AI is a powerful tool, but use care when relying on information provided by AI or sending your confidential information to an AI system.

McConnell Law Firm is committed to the ethical use of technology in our practice, including AI. We look for ways to streamline the practice of law and make it more efficient so we can provide high-quality service to our clients. The time and cost savings we gain through technology, used within the bounds of the law and legal ethics, allow us to focus our energy on providing higher-quality legal advice.

Schedule a Consultation  →  Book a no-cost 30-minute initial consultation with McConnell Law Firm PC to discuss timing and strategy to protect your business and innovations.

This article is for informational purposes only and does not constitute legal advice. Consult a licensed attorney for guidance specific to your intellectual property and business circumstances.

Robert McConnell is the Principal Attorney of McConnell Law Firm PC, with more than 20 years of startup and legal experience in both Silicon Valley and Los Angeles.