DoNotPay Do Not Sign

GPT-3 Chatbot Lawyer DoNotPay Claims Paraphrasing Fine Print Somehow Reveals Hidden Legal Terms

AI legal advice platform DoNotPay released a new version of its Do Not Sign tool for analyzing the long and tedious terms and conditions text that nearly everyone agrees to without reading. The “robot lawyer” platform has offered the service since 2019, but the latest iteration is DoNotPay’s first to employ OpenAI’s GPT-3 generative AI model. DoNotPay CEO Joshua Browder claims the AI can spot red flags and anything unusual in the thicket of legalese. Judging from the output, GPT-3 finds legal jargon just as dull as any human would, offering unimaginative paraphrases of nearly every clause.

GPT-3 ToS

The GPT-3 version of Do Not Sign processes the text of the legal terms and services common to most modern products, particularly technology and software. The AI is supposedly trained to spot red flags and unusual clauses that might be a problem for consumers. It then extracts those phrases and explains what they mean, as seen in the above image. While that may be potentially valuable, the double handful of examples I fed into Do Not Sign all populated the page with details that are hardly surprising, though they can be inconvenient. The Amazon Alexa and Google Assistant terms and conditions had nearly identical results from DoNotPay. While a tech company’s ability to share personal data or shut down a product feature without warning is annoying, it’s also entirely standard, despite the overlong explanations DoNotPay attaches to each clause.

Court Adjourned

Do Not Sign and other consumer rights-centered projects have abruptly become the central focus for DoNotPay this week after Browder curtailed his plan to send the robot lawyer to a courtroom following warnings from “State Bar prosecutors” that an AI listening to a court case and generating responses for a defendant via wireless earbuds might break the law. Browder had offered a million dollars to anyone who would wear AirPods to the Supreme Court and repeat what the AI suggested. The Supreme Court and many other courtrooms ban electronics, though, and accessibility allowances would not likely apply in this case.

That might not have been enough for DoNotPay to pivot away from legal processing entirely, but a viral Twitter thread turned Techdirt article by Kathryn Tewson provoked skepticism over whether AI had much of a role in producing complex legal documents for DoNotPay. Tewson attempted to generate a defamation demand letter and was told it would take an hour, while a divorce settlement agreement was promised in eight hours (a timeframe noted by Voicebot for the same service). That’s very slow for a generative AI system, and they still weren’t ready by then. DoNotPay produced a small claims lawsuit filing much more quickly but included clauses unaddressed in the initial set-up and looked like a previously completed document with some of the proper nouns replaced. All three services are no longer available on DoNotPay. Tewson ended up getting a refund and was blocked on Twitter by Browder.

DoNotPay’s decision doesn’t mean lawyers have no use for GPT-3 and generative AI. Legal technology developer Rally now offers an AI assistant called Spellbook that uses GPT-3 to compose contracts. Spellbook generates suggestions on improving the language in a contract, summarizing the document to be understandable even to a child, and offering ideas for where negotiation could happen. The same goes for legal tech startup Lexion’s AI Contract Assist, a Microsoft Word plugin, and AI legal assistant startup Harvey, who raised $5 million in a funding round led by the OpenAI Startup Fund to answer questions and complete tasks for lawyers.

Legal Tech Firm Rally Showcases GPT-3 AI Contract Writing Assistant Spellbook

Legal Tech Startup Lexion Tasks GPT-3 to Help Draft Contracts in Microsoft Word

AI Legal Assistant Startup Harvey Raises $5M Led by OpenAI