Artificial intelligence (AI) is not new. Documents generated by do-it-yourself online legal services are not new. What is new is the widespread use of AI, especially by nonlawyers, to ask legal questions and generate or analyze legal documents, write code, and generate works of art or authorship. Oh, and lawyers are using it, too.
This area is evolving at internet speed. We thought we might pause for a moment (time stamp: April 2026) to take stock of several things:
- AI is great for some things, but for now the output needs to be checked and double-checked for accuracy and completeness. AI can help summarize and organize documents and assist with first drafts, but some spectacular misfires have occurred from AI “hallucinations” of case law, statutes, and other materials. And AI has no knowledge of the real world or of human relationships or human interactions. The jury is out on whether it is saving any time (or legal spend) at the moment, although it is improving rapidly.
- Be aware that AI-generated computer code, texts, images, music, inventions, etc., are all unprotected under current agency interpretations of copyright and patent laws. Only humans can be authors or inventors. This is likely to become a huge area for litigation as lawyers grapple with limited causes of action when what were traditionally company assets wind up in the wrong hands.
- Materials fed into some AI engines are used for global training and may reappear whenever a prompt (entered by anyone, anywhere) calls it out. Other AI engines have more limited deployment, such that they do not feed the general model. Results stay within the company. It is critical for managers to know which one is being used. However, there is still an issue with the application of copyright and patent laws, or the lack thereof.
- Similarly, secrets should not be spilled by anyone to these open AI systems. As tempting as it may be to get legal advice or therapy from a chatbot, it is not a relationship subject to any privilege. Just as Google searches have been discoverable, so are AI prompts and output.
- Company policies and procedures concerning the use of AI from any device (including personal phones) should be reviewed and updated and additional training of employees and contractors will be required. This should be an ongoing effort alongside cybersecurity training.
- Speaking of cybersecurity, AI has increased the velocity of the arms race. Spoofing of emails is nothing compared to the spoofing of voices and images of executives in deepfakes that misdirect payments. AI agents are also getting very good at finding vulnerabilities in electronic systems.
- Clients who present AI-generated or -reviewed materials to their lawyer should help set expectations about the use of the materials, which may be great, but they will need to be reviewed carefully and their origin may be questioned.
- Lawyers have training and experience in the real world with drafting and negotiating contracts and other documents. AI is great for pointing out issues but does not know how real people may react. We must all remember that AI sounds smart, but it is still very young and is only simulating the things humans learn over time.
For at least a few years, there will be what some are calling a “verification tax” on generative AI. This “tax” is the extra time lawyers will need to spend to manage the uploading of information and to check everything that the AI program suggests, cites, or insists is missing. The human lawyer is the first and only line of defense against AI slop.
To learn more, please contact the author or any member of the Intellectual Property Group at McCarter & English.