Why You Shouldn’t Rely on AI to Review Your Contract

Reaching the contract phase is an exciting step in your family-building journey. We know opening a 40-page legal document for the first time can feel a bit overwhelming. It’s completely normal to have questions about the terms that directly impact you, like insurance, timelines, or financial arrangements. When clients ask us questions about the first draft we sent, we are here to provide clear answers to help them stay informed, as we know this may be their first time navigating a legal agreement of this scope.

Recently, though, we have started receiving more emails with long lists of questions organized in neat bullet points. At first glance, they seem thoughtful. As we read further, however, many of the questions turn out to focus on clauses that have little to do with the clients’ actual situation. In cases like this, we can tell that clients have run their contract through an AI chatbot and then forwarded us the list of questions it generated.

We understand the appeal of AI. It's tempting to hand a dense document to a free tool that can flag ten potential issues in seconds instead of reading every line yourself. Nevertheless, from our years of experience drafting and reviewing these contracts, we’ve found that the AI tool clients use to engage with their contracts often ends up creating risks and confusion that it’s supposed to help avoid. In this blog, we would like to discuss three major issues with running your contract through AI. 

1. Confidentiality Risk: Where Does Your Contract Actually Go?

What Happens Behind the Scenes When You Paste a Draft

The first issue is one most people never think about: the moment you send your contract draft to an AI chatbot, where does it go? When you paste your contract into ChatGPT or another GenAI tool, it doesn’t “read” the document the way a person does. The text is broken into smaller units called tokens and converted into numerical representations that the model uses to generate a response. In most GenAI systems, your input is then sent to the provider’s servers for processing. 

Although interacting with an AI chatbot may feel like a private conversation, the contract leaves your device and enters someone else’s infrastructure. It is then handled under the provider's privacy policy and terms of service. Depending on the platform, it may be temporarily stored, logged for security purposes, reviewed by humans for quality assurance, or used to train future models. OpenAI’s own data-usage policy, for example, states that content submitted to its consumer services, including ChatGPT, may be used to train its models unless the user opts out. In some cases, fragments of your contract may appear in what the AI generates for other users. From the moment you upload it, you are no longer the one deciding how that information is handled. This is where confidentiality risks appear. 

Losing Attorney-Client Privilege

When thinking of the term “confidentiality”, some people understand it as a promise that the information will not be actively shared with any individuals outside an authorized circle. In everyday practice, however, it consists of more layers. Confidentiality, the obligation to protect private or sensitive information from unauthorized disclosure, also involves a set of strict protocols regarding how such information is handled. In a law firm, that responsibility is built into everyday practice: who can open a file, where documents are stored, and how they are shared. 

These protective boundaries exist precisely in line with the legal rule of attorney-client privilege, codified in California Evidence Code section 954, which protects communications between you and your attorney so that there’s always a guarantee that the sensitive information you disclose to your attorney will remain confidential, and you will not be discouraged from sharing facts that are deeply personal or critical to your case. However, the privilege can be broken if you share such information with unauthorized third parties, which is exactly what you risk doing when you upload your contract to an AI chat. 

The risk is not merely theoretical. Just in 2026, in United States v. Heppner (S.D.N.Y. Feb. 10, 2026), a federal court ruled that a criminal defendant's litigation strategy questions and interactions with a public AI chatbot (Anthropic’s Claude), were discoverable by the government and used as evidence against him, was not protected by attorney-client privilege. Even though the defendant was analyzing information received from his lawyers and later shared the AI’s output back with them, the court held that handing the information over to a third-party AI provider destroyed the expectation of privacy.

For attorney-client privilege to apply, three conditions must be met: It must be a communication made between a privileged person and an attorney, in confidence, and for the purpose of seeking or receiving legal advice. Pasting sensitive contracts that are provided by your attorney into a public AI chatbot potentially breaks these conditions and therefore waives the legal protections that would have otherwise kept your private details secure. 

Although the legal consequences of sharing this information depend on the circumstances and the tool, the surest way to preserve attorney client privilege is not to hand the communication to a third party at all.

Risking the Violation of Confidentiality Clauses

Additionally, many agreements also contain confidentiality or privacy clauses that limit what information about the case can be shared and with whom. In surrogacy and donation agreements, this is especially the case, as disclosing identifying information to an unauthorized third party can jeopardize the privacy and security of everyone involved, including the future child. The contract often contains parties' full names, addresses, email addresses, compensation details, and family planning decisions, all of which are highly sensitive. 

Uploading these drafts to a public AI platform can potentially be considered a violation of the confidentiality clause. Human reviewers routinely read chat logs for quality control. Worse, chatbots can be tricked with prompts to remove barriers to providing this info to irrelevant users. Operating inside an opaque black box, these hidden data flows leave you with zero control over who ultimately sees your and the other party’s personal information. 

Although the wording of each contract regarding confidentiality can be different, and it is not always the case that an upload triggers an automatic breach, protecting your private data is always the safest approach. Without reading the contract itself first and understanding the clauses carefully, sharing it with an AI on a whim can put you at significant legal risk.

2. Accuracy Risk: When AI Hallucinates

Word Prediction ≠ Legal Reasoning

If the confidentiality risk is about the action of uploading your data, the second risk is about the output and what can go wrong if one over-relies on the response provided. When you ask a chatbot to analyze a contract, you are risking what is commonly called AI "hallucination." This happens when a model produces answers that sound confident and convincing but are entirely detached from reality. 

This is not uncommon in the legal field. In the landmark 2023 case Mata v. Avianca, Inc., 678 F.Supp.3d 443 (S.D.N.Y. 2023), a federal court sanctioned an attorney who submitted a brief filled with fake case citations entirely hallucinated by ChatGPT. More recently, in Noland v. Land of the Free L.P. (Cal. Ct. App. Sept. 12, 2025), a California appellate court issued a $10,000 fine to a lawyer who similarly submitted a brief containing fabricated AI-generated citations. In fact, the problem has become so widespread that a public database maintained by legal researcher Damien Charlotin now tracks it, documenting nearly 1,800 court decisions worldwide in which a party relied on AI-hallucinated material. If AI models can so easily mislead practicing attorneys, it is even more dangerous for our clients without legal backgrounds to rely on them.

How does AI hallucination happen? Unlike an attorney who conducts legal research by analyzing case-specific facts and looking up the current statutes or cases that apply, AI generates answers by identifying patterns in large amounts of training data and statistically predicting the most likely sequence of words. 

Unreliable Web Data and Losing Legal Nuances

The issue with this mechanism is that much of the training data AI relies on is not legal text at all, but was scraped from web sources like Wikipedia and Reddit. Therefore, having a general-purpose AI model review your contract is akin to asking a web forum for legal advice. Even when it does use data from public court records and government websites, these inputs may be outdated or completely irrelevant to your situation. 

What makes this even more dangerous is that these models are tuned to sound persuasive and complete. When AI gives answers based on a blend of internet chatter and outdated legal fragments, it can still pass as an authoritative opinion. Sometimes, when a question falls outside what the model actually knows, instead of saying "I'm not sure.”, it fills the gap with what is most “likely” the answer that the user looks for, whereas the answer may be factually incorrect or unverified. This shifts the burden entirely onto the user, who usually lacks the specific legal training, to attempt to distinguish true statements from false ones.

While some might believe that gaining some broad legal knowledge from AI is still better than having none at all, in surrogacy law, broad legal advice is sometimes worse than useless. Surrogacy in the United States is not regulated at the federal level; it varies state by state. A clause that is fully enforceable in California under Family Code sections 7960 through 7962, may be void, or handled entirely differently, in another state. If you ask an AI a general question about such a clause, it may answer from general family law content on the internet, rather than the specific statute that actually governs the agreement. These nuanced variables can hardly be accurately handled through AI’s guesswork.

How AI Feedback Loops Escalate Your Stress

Besides drawing from unreliable inputs, there is a second way AI models can hallucinate. AI models are built to be responsive to the person asking, and when prompted repeatedly, they show sycophantic behavior, that is, a tendency to mirror the view of the user. Therefore, when someone, feeling uncertain about a complex legal agreement, turns to an AI to help understand, it can inadvertently trigger an escalating feedback loop. The model can potentially read a request like "Is this clause unfair to me?" as a command to "find flaws," thereby hallucinating risks that are completely ungrounded in the text of the contract. Because these tools present information so persuasively, it is natural for a user to take these warnings seriously and ask follow-up questions, which, occasionally, prompts the AI to generate very detached speculation. 

Very often, we notice AI-generated questions in client emails because the issues they raise are either irrelevant to the contract itself or already clearly addressed in the contract language. These detours stall what could have been a more straightforward review. In a family-building journey, which is often mixed with both joy and anxiety, we see AI often end up amplifying the stress rather than providing clarity or peace of mind that the families look for. 

3. Individualized Representation: More than Information

Another reason why we discourage the use of AI, beyond the risk of breach of confidentiality and unreliability in interpreting legal text, is that it cannot replicate the individual representation an experienced attorney provides. Representation involves much more than just explaining legal concepts to clients or retrieving information from clients to be entered into an agreement. What distinguishes representation from advice generated from AI is the legal accountability and professional judgment of an attorney and their deep understanding of the entire professional ecosystem managing a surrogacy or a donation journey.

Professional Accountability and Legal Ethics

First, when an attorney is retained to draft or review your surrogacy or donation agreement, a formal relationship is established under strict professional and ethical rules. An attorney can be sued for professional malpractice, and therefore carries a direct legal obligation to ensure the advice they provide is accurate. Every state enforces its own rules of professional conduct for attorneys. In California, the Rules of Professional Conduct impose, among other things, the duty to possess competence in the relevant field of law (rule 1.1), the duty of confidentiality (rule 1.6), and the duty of avoiding conflicts of interest (rule 1.7), and an attorney's license is regulated by the jurisdiction in which they practice. 

Under these rules, attorneys must be responsible for their advice and the judgment calls they make on their client’s behalf. An AI model, by contrast, is not a licensed counsel. If an AI argues for unconscionable medical decision provisions, which subsequently causes the contract negotiations to be terminated, it cannot be held accountable for the consequences. 

Why Human Judgments Matter in Your Contract Review

Many people might still find it tempting to use AI as a “backup” tool to check the agreement. In practice, however, we find this approach frequently introduces more confusion. When provided with an entire agreement to unpack, general-purpose language models tend to treat every provision on the page with equal weight and generate exhaustive lists of hypothetical risks or irrelevant legal theories. Without clear directions, they are often incapable of distinguishing clauses that reflect standard statutory language and the more high-stakes clauses that truly require a client’s attention.

In our contract review process, the most valuable conversations usually go beyond simply explaining what the agreement says. Clients often ask about how certain provisions may affect them in practice or request changes based on their own needs, such as additional medical screening requirements, adjustments to compensation or benefits, or specific expectations during the pregnancy. These questions cannot be answered by simply reviewing the language of the agreement. They require the attorney’s experience with similar cases and a deep understanding of the client’s individual situation, the negotiation process, and the agency involved. These are precisely the human judgments that our attorneys are best fit to make using their expertise. 

What AI Misses in Real-World Negotiation

Some may argue that AI has become more personalized because many AI tools can now remember information shared by users across different conversations. While this feature may make AI responses feel more tailored and even adept in suggesting potential contractual provisions, using it does not necessarily facilitate the process of finalizing a contract. Knowing what the client wants to request as it comes to a surrogacy or donation agreement represents merely the baseline of a legal process. 

Between when the first contract draft is provided and when a contract is finalized, the negotiation process demands a sophisticated understanding of the other party in the agreement and their representation. Experienced legal counsels always use their knowledge of both parties and prior experience to strategically advise clients on the specific conditions they may not want to yield on versus those where they can safely offer a middle ground. Despite some knowledge shared by the user, an AI model remains entirely blind to these tactical issues as it lacks much information about all the stakeholders involved. Therefore, they cannot predict how the other party may react to a demand, nor can they consistently inform the user of the long-term risks embedded in a seemingly minor compromise. More often than not, relying on AI to analyze these fluid variables frequently introduces friction and distractions. 

Embracing Innovation While Protecting Your Journey

None of these is to discourage clients from seeking additional resources when reviewing a contract. We understand why many clients may turn to AI before speaking with their attorney. Surrogacy or donation agreements are lengthy and include unfamiliar legal terminology, and this can be especially challenging for international intended parents who may be navigating both a foreign legal system and a second language. Some clients may feel more comfortable using AI as a low-pressure way to organize their thoughts, explore questions they may be hesitant to ask, or gain confidence in their own understanding before discussing the agreement with counsel. 

However, using AI for general background, like understanding the surrogacy process, is very different from feeding it an entire contract draft in hopes of receiving legal analysis. From a practitioner's perspective, uploading agreements to general-purpose models creates unintentional privacy risks, adds unintended layers of confusion to the collaborative review process between the client and the attorney, and pulls focus away from the key information within the agreement. If you wish to request a summary of a surrogacy or donation agreement, you can ask, and we can provide one. We can utilize professional versions of AI where we review the policies to ensure they will not be subject to third-party disclosure. 

AI is changing the legal industry quickly, and we welcome these changes. Our firm actively uses new technology in our own practice to best serve our clients. With generative AI, we even built our own IVF calculator app to help intended parents and surrogates to track the pregnancy timeline and due date after an IVF procedure. We imagine that it would shape the legal industry in more ways than we previously anticipated, and we are not surprised that our clients are also finding the tool compelling.

Yet, as AI unlocks unprecedented access to information, we want to remind our clients that information is still not a substitute for decades of legal experience. You do not have to turn to an AI-generated checklist for peace of mind, as it still remains a risky and inaccurate tool for a legal journey that largely requires human judgment.  When your draft arrives, we encourage you to read it yourself through the lens of your family’s real-world priorities and speak to us honestly about what’s on your mind. You can trust us to take care of the legal analysis and the tough negotiations so you can focus on the journey ahead.

Let’s Protect Your Family-Building Journey Together.

Have questions about your draft agreement? Tsong Law Group provides dedicated, strategic contract review for intended parents, surrogates, and donors across California, New York, Illinois, Washington, Arizona, and Oklahoma, as well as international intended parents building their families in the U.S. Contact us at info@tsonglaw.com or 877-992-3678 to schedule a consultation with an experienced attorney.

This article is provided for informational and educational purposes only and does not constitute legal advice or establish an attorney-client relationship. Every contract and family-building journey is unique; please consult a qualified legal professional regarding your specific legal needs and state laws.