# Can You Use ChatGPT For Redlining Legal Contracts? Can ChatGPT redline contracts accurately, and can you ethically use it? We break down what it can and can’t do, and why general AI falls short for legal work. Explore the risks of hallucinations, legal nuance gaps, and lack of integration, plus how legal-specific AI tools like Gavel Exec fill in the gaps to support contract review and negotiation. By the team at Gavel Published: Jul 16, 2025 | Last updated: Jul 24, 2025 ## What ChatGPT Can Actually Do in Contract Redlining ChatGPT, as a large language model, is good at: - Rewriting clauses in simpler or more readable language - Flagging inconsistent terminology (e.g., “Agreement” vs. “agreement”) - Suggesting plausible alternative phrasing - Explaining legal text in plain English - Providing structural outlines of contracts - Creating first drafts of boilerplate language Think of it like a hyper-efficient intern; it can generate text quickly and reasonably well, but it doesn’t know what matters in your deal or what’s acceptable to your client. That distinction is critical, and ethically consequential. ## Ethical and Legal Requirements for Using AI in Legal Work State bars and ethics committees have begun to weigh in on how tools like ChatGPT can be used in practice. Here’s a summary of where things stand: ### Bottom line Lawyers can use generative AI, but only if they ensure accuracy, protect confidential information, and disclose usage appropriately where necessary. The burden is on the lawyer, not the AI provider, to comply with professional standards. ## Where ChatGPT Falls Short, and Why That Matters Ethically ### **1. ⚠️ Lack of Legal Judgment or Context** ChatGPT doesn’t understand your client’s objectives, deal dynamics, or jurisdictional requirements. It predicts text that _sounds_ right but doesn’t evaluate legal impact. It may: - Treat a major shift in liability like a minor wording change - Suggest removing a clause that protects your client - Miss jurisdiction-specific requirements (e.g., indemnity limits in California vs. Delaware) #### Ethical Implication Under Rule 1.1 (Competence), you must supervise any AI output and ensure it meets the legal and strategic needs of the matter. ### **2. 🧠 Hallucination Risk and Factual Inaccuracy** ChatGPT is known to "hallucinate", fabricating legal clauses, case citations, or regulatory references. In the Mata v. Avianca case, lawyers submitted AI-generated cases that didn’t exist. That’s not just embarrassing; it led to sanctions. #### Ethical Implication Rule 3.3 (Candor to the Tribunal) and Rule 4.1 (Truthfulness) prohibit false statements, even if generated by AI. You’re responsible for verifying all facts and legal authority the AI presents. ### **3. 🔒 Client Confidentiality Violations** The public version of ChatGPT operates on shared infrastructure and may retain user inputs for training unless certain settings are enabled. Inputting confidential client data into such a system may: - Waive attorney-client privilege - Breach Rule 1.6 (Confidentiality) - Violate your firm’s information security policies #### Risk mitigation Only use platforms that offer **Zero Data Retention**, encrypt all data, and sign data protection agreements. Many general AI tools (including ChatGPT) do not meet this bar out-of-the-box. ### **4. 🚫 No Awareness of Firm Playbooks or Negotiation Strategy** ChatGPT doesn’t know your fallback positions, clause libraries, or preferred language. It won’t know: - That your firm _always_ includes a cap on liability - That arbitration is non-negotiable in your SaaS agreements - That a specific indemnification term is tied to past litigation exposure #### Ethical Implication Rule 1.1 and Rule 1.3 (Diligence) require you to apply your firm’s legal standards and client preferences. If AI output contradicts your policies, and you don’t catch it, you may compromise your client’s interests. ### **5. 🧾 Billing and Disclosure Issues** If AI saves time on document review, how should you bill for that work? Ethically, you must: - **Avoid inflating hours** for work partially or fully completed by AI - **Explain AI-assisted work clearly** if charging based on value or fixed fees - **Disclose AI use** if required by court rules, local practice, or client engagement letters Billing transparency is governed by Rule 1.5 (Fees) and Rule 7.1 (Communication About Services). Some bar opinions suggest that disclosure may also be needed under Rule 5.3, since generative AI is considered nonlawyer assistance. ## Why Human Oversight Is Legally and Ethically Required Even if AI does 90% of the redline, you are responsible for 100% of it. That’s not just good practice; it’s what ethics rules demand. Every AI-generated edit must be: - **Reviewed** by a licensed attorney - **Assessed** for legal accuracy, client alignment, and jurisdictional fit - **Evaluated** in light of negotiation leverage and business context - **Adjusted** for tone, diplomacy, and presentation style You can’t delegate legal judgment to an algorithm. You can’t assume the AI “knows what it’s doing.” And if something goes wrong, “the software told me to” is not a defense. ## What Legal-Specific AI Tools Like Gavel Exec Do Differently To address these risks, some tools have been designed specifically for lawyers, tools like **Gavel Exec**. Unlike general-purpose AI, Gavel Exec complies with ethical and practical requirements out of the box. ### Key Guidelines for Ethical AI Use in Contract Review If you're considering AI tools like ChatGPT for legal work, follow these **minimum best practices**: **DO:** - Use AI only with full human supervision - Review every AI output before it reaches a client or counterparty - Protect client data with compliant tools (e.g., zero-retention, secure cloud) - Clearly disclose AI-assisted work when required - Train your team on AI ethics and internal usage policies **DON’T:** - Input sensitive client data into public ChatGPT - Rely on AI for legal advice or high-risk edits - Submit AI-generated redlines without thorough review - Assume that because it “sounds good,” it’s legally safe - Overbill for work done primarily by an AI ## Final Thoughts: ChatGPT is a Tool, Not a Legal Colleague ChatGPT can be helpful for brainstorming, rewriting, and flagging surface-level issues. But it cannot replace legal expertise, ethical responsibility, or judgment. Tools like Gavel Exec go further, integrating legal workflows, protecting client data, enforcing your playbook, and making AI a real asset to your team. But even with legal-specific AI, you must stay in the driver’s seat. Ethics rules don’t change just because the technology does. Use AI. Use it smartly. And above all, use it ethically.