It feels like a private conversation. You type a question like you would to a human, get a quick response (sometimes with a compliment, “Terrific question!”) and move on. But when you're using an AI chat tool, especially with legal work, that sense of privacy can be misleading.
For attorneys, protecting various privileges, confidentiality, and privacy are more than a binary setting or preference. Confidentiality, privilege and client trust are not abstract principles. Preventing client materials against disclosure is a daily responsibility as a professional ethical obligation. As AI systems become more integrated into legal workflows, it raises an urgent question: Who else might see the data you enter into that chat window? Is the answer similar to thinking through “who might read this email?” and deleting certain text before clicking send just in case the email gets forwarded?
The answer might not be as straightforward as you think. Let’s take a closer look at how AI platforms handle your information, why that matters for legal professionals and what steps you can take to protect sensitive data.
For this post, “input” means the documents/media you upload and/or your prompts, and “output” means the content generated by an AI tool in response to your input.
Free vs. Business AI Tools
Not all AI tools treat your information the same way. To most users, the free and paid (business) versions of an AI tool look identical. The difference between the two goes beyond features, they come with different licenses. The licenses determine how data is handled, where it goes and who might have access to it in the future. Here are the notable differences:
- Free AI tools are built for consumer use and, in exchange for cost-free access, may (probably) collect or review your input, offering little to no assurance that sensitive legal data will remain private.
- Business-specific platforms are designed for professional environments. They may or may not come with additional features and capabilities, but they typically come with formal data handling policies, stronger encryption and enterprise-level safeguards spelled out in their licenses. Some business-licensed AI tools allow users to opt out of model training. Others offer private deployment options with clear access controls and usage logs. Some platforms offer both free and paid versions, which may have distinct data handling practices and privacy protections.
For outside counsel in particular, this distinction matters.
Using an AI tool under a free license may expose client data in ways that breach confidentiality or violate professional ethic rules. If a problem occurs in these scenarios, the liability falls squarely on the person or firm that (inadvertently) shared the data; the AI provider disclaims liability for pretty much everything. Even if no immediate confidentiality breach occurs, the risk to client trust and professional reputation is significant.
If you’re not sure whether your AI tool is truly private, it probably isn't.
Attorneys must treat AI platforms like any other vendor handling privileged and/or confidential material. That means reviewing terms of service, asking the right questions and choosing tools that match the privacy obligations of legal practice.
Privacy is a primary concern when using AI, but this tool also comes with input and output risks, which Zero Day Law discusses in this recent blog post.
Behind the Curtain: How AI Companies Actually Handle Your Data
Though most users focus on the front-end output, behind the scenes there are several areas of concern if legal professionals use artificial intelligence to aid in their legal work:
- AI chats may not be completely private and fully automated: Human reviewers and developers or backend diagnostic tools may access conversation input, output and logs, and take screenshots or session recordings to improve system performance or monitor reliability, which can raise serious privacy concerns if not contracted properly.
- Data retention varies among AI providers: Some platforms keep inputs for a month, others indefinitely. Deletion options exist, but are often hard to find. Free versions typically log everything by default.
- Large Language Models are constantly becoming smarter through training: Unless you're using a business version with strict controls, your inputs may be used to train the model. That means client data or legal drafts could be reused without your knowledge.
For attorneys, these factors provide clear risks. A platform that does not offer strong, enforceable privacy terms and controls should never be used for any sensitive work.
Learn more about the balance between privacy regulations and AI use in this recent blog post.
Understanding Current AI Agreements
Before sharing any confidential information with an AI platform, take time to review the terms that control how your data is handled. These details are often buried in user agreements and privacy policies. Many users overlook them, but they carry real legal consequences. Businesses may assume they’ll negotiate a license for a Business copy, but often the generic business license online applies.
The following are current links to the terms of service or user agreements for popular platforms:
These documents define what the provider can do with your inputs and how long they can keep them. Some include clauses that allow internal staff to view user content. Others grant broad rights to reuse data for research or system improvements.
What to Look For in AI User Agreements
As you review these documents, focus on several key points:
- How the provider defines “user data” and “content”
- Whether and where the company can review, store, or use your inputs for training its models and/or a model solely for your use
- What the company can and can’t do with your inputs and outputs
- How long the data is retained
- Whether they share any of it with third parties or affiliates
- Whether they own or license your inputs and outputs
But user beware; terms change often. A platform that was safe to use last quarter may have adopted new policies without notice. Make it a regular habit to check for updates, especially if your firm relies on these tools for any part of its workflow.
Explore more about the connection between AI and cybersecurity privacy.
A Practical Protocol for Legal AI Usage
Law firms can’t afford to approach AI use casually. A formal protocol protects your clients, your team, and your liability exposure. This advice is not entirely different from how we’d advise lawyers to protect data in Sharepoint, Google Workspace or email clients. But with AI, people tend to treat it differently and not follow long-standing paradigms. Zero Day Law recommends a framework that includes:
- A Platform Review: Free tools generally offer little control over how your inputs and outputs are used. Each enterprise grade platform has their own terms and they are not created equally.
- Internal Information Categories: Silo information categories, tagging data types suitable for AI assistance and other categories of restricted information. When in doubt, redact or revise or avoid AI use entirely for these scenarios. Avoid using AI tools as another place to store documents.
- Comprehensive training for your entire team on appropriate AI boundaries: Put clear guidelines in writing and set up a process to document how AI is used for drafting, productivity or research. Review and revise these practices as technology and legal standards change.
- Data minimization: Minimize the amount of data uploaded or inputted into the AI tool and document which data is allowed to be used in an AI tool. Remove input and outputs from the AI tool when they are no longer needed and ensure this is technically possible within the platform beforehand.
“Treat every AI chat like a shared workspace, not a private notebook. If you wouldn’t say it in front of opposing counsel, don’t type it into a chatbot.”
Without a strong structure in place, even careful use can expose your firm to unnecessary risk. Clear, easy-to-understand guidelines will help protect your team and your clients.
Tactical AI Privacy Protection Strategies
Even with secure platforms and sound policies, attorneys should take extra steps to avoid exposing sensitive material:
- Scrub identifiable information, replacing it with initials or generic terms.
- Redact locations or timelines that could link content back to a specific matter.
- Scrub any details that are unnecessary for your work; they can often lead to revealing more than you might think.
- Adding intentional “noise” can also be effective. This approach blends real facts with hypothetical scenarios, even introducing deliberate errors that can be corrected later. This approach reduces the ability to trace information back to a particular person or situation.
Enterprise versions often include access control, disabled logging and training features and other technical safeguards. For firms dealing with highly regulated information, on-premise LLM deployments and AI tools will keep all data within a contained network.
When Legal Professionals Should Avoid AI Entirely
There are times when no amount of caution or technical protection is enough. Some matters should never be fed into an AI system, including:
- Privileged conversations with clients
- Confidential information related to ongoing investigations or litigation
- Draft regulatory filings or compliance-related legal analysis
If the content is sensitive enough to warrant a locked filing cabinet or secure server, it should not be processed through a third-party AI tool.
Balancing AI Innovation with Legal Protection
AI can speed up research, streamline document review and generate new ideas. But the legal profession has obligations that go beyond a general obligation to be efficient for clients. Confidentiality, client trust and ethical compliance must come first.
AI does not need to be avoided entirely in legal settings. A clear use plan will properly protect your firm and your clients. Assess the platforms you rely on. Classify what information is safe to use. Train your team, document your practices and revise your approach regularly.
Protecting client confidentiality in the age of AI starts with knowing how your tools really work. Not just how their websites say they work! If your firm is using AI or considering it, a clear, enforceable plan will help protect sensitive information. The risks are real, but with the right protocols, privacy and performance can coexist.
Learn more about Zero Day Law’s AI Privacy and Security Solutions here.
Our team can help your firm evaluate current tools, draft policy language and build a confidentiality-first AI workflow. Contact us today to start the conversation or request a consultation. Zero Day Law is here to help you keep security, compliance and client trust at the center of your AI-use strategy, whether that's helping with policy, terms or just demystifying legal gray areas that come with emerging tools and technologies.