CoCounsel leverages the latest breakthroughs in A.I. technology in the form of GPT-4, which we have tailored to perform a wide range of sophisticated legal tasks. CoCounsel runs on a large language model, but it differs in a number of critical ways from ChatGPT and similar “chatbots” that use publicly-available large language models. These differences, summarized below, make CoCounsel a secure platform that attorneys can trust for accurate, reliable information.
Specific legal capabilities: CoCounsel can complete substantive legal tasks in common litigation and transactional practice areas. By contrast, ChatGPT and similar chatbots that run on public large language models were not developed for attorneys and are not intended to be used in the practice of law.
Extensive testing: CoCounsel has been rigorously tested by Casetext’s internal team of litigators and transactional attorneys, who work closely with Casetext’s AI engineers to ensure that CoCounsel’s answers are accurate and can be used by attorneys in their daily practices. All of CoCounsel’s capabilities have been tested with thousands of internal tests. After clearing the internal testing phase, CoCounsel's capabilities were then released to our beta testing group, which was composed of hundreds of attorneys from elite boutique and large law firms, for additional testing before finally being released to the public.
Guardrails against “hallucinated” authorities: CoCounsel only performs legal research and document review from real, existing sources. For example, CoCounsel will only write a legal research memo after finding and reading relevant cases, statutes, and regulations. By contrast, ChatGPT and other general use AI models do not have these guardrails in place and can therefore make up or “hallucinate” legal authorities that do not actually exist.
Daily updates: CoCounsel’s legal research application pulls from Casetext’s database of primary law, which is updated daily. No such regular, domain-specific updates exist for ChatGPT or other large language models, which can generate language based on years-old training data.
Know whether you are citing good law: All cases cited by CoCounsel are checked by Casetext’s “SmartCite” citator to determine whether those authorities are still good law. No such protections exist with ChatGPT or similar public AI models.
Robust security protocols: Casetext is deeply committed to security and privacy and has implemented rigorous security protocols to maintain the confidentiality of your data. All data you provide to Casetext when you use CoCounsel, including any and all information and documents you choose to upload to CoCounsel, are encrypted and protected in accordance with Casetext’s comprehensive security program. None of your information or documents are used to train the CoCounsel platform or the underlying large language model. These security measures may not exist with other generative AI applications, where your data may be accessed and used by others.