PlaintiffStack vs ChatGPT

Why so many lawyers use ChatGPT

General AI assistants are already part of plaintiff practice. In the 8am 2026 Legal Industry Report, among personal injury respondents, 66% said they use general tools like ChatGPT or Claude and 44% use legal-specific AI. 78% said they use AI tools personally even where their firm has no official policy. The same respondents named accuracy (70%), ethics (69%) and data security (68%) as the top concerns slowing firm-wide adoption.

That makes sense. ChatGPT is fast, inexpensive, good with language, and available to anyone. For tasks that involve no client information, it is often the right tool.

Confidentiality and client data

The questions start when client information goes in. On consumer plans (Free and paid individual plans), OpenAI uses conversations to improve its models by default unless the user turns that off in Data Controls. On ChatGPT Business, Enterprise and the API, OpenAI says it doesn't train on inputs or outputs by default.

Medical records add HIPAA. OpenAI's HIPAA guide says its services may not be used with protected health information without a BAA. It offers BAAs for the API and for ChatGPT Enterprise and Edu configured for it, and launched ChatGPT for Healthcare with BAA coverage in January 2026. Consumer plans aren't covered. A paralegal pasting records into a personal account is outside all of that. See our business associate agreement entry and legal AI security checklist.

ABA Formal Opinion 512 and California guidance

The ABA's first formal opinion on generative AI, Formal Opinion 512 (July 29, 2024), applies existing duties to these tools. Lawyers must understand a tool's capabilities and limits, independently verify its output, and protect confidentiality. For self-learning tools that may reuse what's entered, the opinion calls for the client's informed consent before information relating to the representation goes in, and boilerplate in an engagement letter isn't enough. It also covers supervision, communication and fees. The State Bar of California has published its own practical guidance on generative AI. Our guide to Opinion 512 for plaintiff firms goes through each duty.

Citation risk is real

Any language model can produce a confident citation to a case that doesn't exist. In February 2025, a federal court in Wyoming sanctioned three Morgan & Morgan lawyers in a case against Walmart after a motion cited eight nonexistent cases. The drafting lawyer had used the firm's in-house AI platform; he lost his pro hac vice admission and was fined $3,000, and two colleagues who signed the filing were fined $1,000 each. The court declined to sanction the firm, noting the steps it had taken to require independent verification. The tool wasn't ChatGPT, but the lesson applies to every model: check every citation against the source.

Where a case-aware platform differs

PlaintiffStack is a case management system for California personal injury, workers' comp and employment with AI case tools inside it. The difference from a general assistant is less about the model and more about what surrounds it:

• It works from the case. Agents read the case's documents, not whatever someone pasted in, and propose facts, deadlines and tasks.
• It cites your file. Every proposed fact cites the document, page and words it came from, so checking takes a click.
• A person decides. Proposals wait for a person to accept them, and nothing goes to a client, carrier or court until a person sends it.
• The data stays put. A BAA is signed with every firm before client data arrives; everything holding client data runs in one US AWS account, models included, through Amazon Bedrock; firm data never trains models.

PlaintiffStack has limits a general assistant doesn't: it reads only searchable PDFs, Word and text files (no OCR yet), supports California only, and has no case-law research tool. Its SOC 2 Type II is in progress, with no report yet. And its AI can still be wrong, which is why it cites and waits. See Draft and security.

Who should use which

Use ChatGPT, on a business plan with training off, for work without client information. Use a case-aware platform for work inside a client's file, where confidentiality, a BAA and citations to the record matter. Many firms will do both. See also the best AI for personal injury lawyers.