The Risks of Clients Using ChatGPT to Review Legal Documents

The popularity of generative Artificial Intelligence (AI) platforms has surged recently. Clients increasingly use tools like ChatGPT to review documents from their lawyers, attracted by the potential for instant feedback and additional insights. However, this practice carries significant risks and limitations that clients must understand. This article examines the limitations and strategies for managing the risks associated with using generative AI, specifically ChatGPT, to review legal documents. 

Limitations 

Information Accuracy 

A major concern with AI tools is their capacity to generate false responses, known as ‘hallucinations’. When ‘hallucinating’, the tool can fabricate content or cite non-existent cases. Therefore, you must carefully scrutinise and independently verify any content. 

In Roberto Mata v Avianca, a lawyer relied on ChatGPT for research and filed a document referring to several non-existent cases. The court held the lawyer accountable, fining him for submitting misleading information. Although this example involves a lawyer, it highlights the danger of relying on AI for legal accuracy. Clients using ChatGPT might misinterpret or accept inaccurate information, leading to flawed decisions. 

Additionally, ChatGPT’s homepage warns that ‘ChatGPT may produce inaccurate information about people, places, or facts.’ Its knowledge is limited to data available up until September 2021. 

Responses may not be current with laws. This can create legal risks for clients. Clients who rely on old information may be affected. 

Confidentiality 

Using AI tools can inadvertently breach confidentiality, with data supplied by the user potentially entering the public domain through subsequent queries or system vulnerabilities. In March 2023, ChatGPT experienced its first data breach, which exposed various user questions and personal information. AI tools storing or accessing sensitive legal documents reviewed by clients risk third parties accessing this information. 

Privacy 

As with confidentiality, clients must be careful not to input personal data into AI tools. The use and disclosure of personal information are subject to various laws, including Australia's Privacy Act 1988.  

Clients should know that AI products can collect user information. ChatGPT's privacy policy collects IP addresses, browser information, and data on user interactions, which may be shared with third parties. This could lead to unintended sharing of private information. 

Copyright and Intellectual Property

 

The training data used by AI tools likely includes copyrighted material. ChatGPT does not provide source references, posing a significant risk of unintentional copyright violations. Clients may unknowingly incorporate copyrighted material into their reviews or rely on unverified information. 

 

Risk Management 

Managing the risks associated with using AI tools like ChatGPT involves implementing several strategies. By taking these steps, clients can reduce the likelihood of encountering problems and ensure they are using these tools responsibly and effectively. 

Independent Verification 

AI tools are not completely reliable, so all outputs should be carefully checked and treated as supplements rather than definitive sources. Clients should always verify critical legal information from reputable and up-to-date legal sources before relying on it. This means cross-referencing AI-generated content with trusted legal databases, consulting with legal professionals, and keeping abreast of current laws and regulations. 

In practice, this could involve: 

• Double-checking AI outputs: Always cross-check information provided by ChatGPT with reliable legal sources to ensure accuracy. For instance, compare AI-generated case citations with verified legal databases. 
  

• It is important to consult with legal professionals. Even if ChatGPT offers an initial review, clients should still consult their own lawyers. This is necessary for a detailed examination and confirmation of legal issues. 
  

• Stay informed by staying up to date with the latest legal changes. This will ensure that the information you use is current and accurate. 
  

 

Data Protection 

To protect sensitive information, clients should avoid inputting confidential or personally identifiable information into AI tools. Here are some specific steps: 

• Limit input to non-sensitive data: Only use AI tools for general inquiries or non-sensitive data. Avoid sharing detailed personal or confidential information. 
  

• Utilise AI settings: Many AI platforms allow users to opt out of data retention. Make use of these settings to prevent the tool from storing or sharing your data. 
  

• Use anonymisation techniques: When possible, anonymise the data input into AI tools to prevent any potential exposure of sensitive information. 
  

 

Copyright Compliance 

To avoid copyright issues, clients should: 

• Understand copyright laws: Familiarise themselves with copyright regulations to recognise what constitutes a violation. 
  

• Avoid verbatim use of AI responses: Treat AI-generated responses as starting points. Rework and verify the content independently before using it in any official documents. 
  

• Cite sources appropriately: If using AI-generated content that references external sources, ensures proper attribution and compliance with copyright laws. 
  

 

Privacy Compliance 

Ensuring privacy compliance involves adhering to relevant privacy laws and regulations. This includes: 

• Understanding privacy laws: Be aware of the privacy regulations applicable in your jurisdiction, such as Australia's Privacy Act 1988. 
  

• Minimise data sharing: Avoid unnecessary sharing of personal data with AI tools. Input only what is necessary for obtaining a useful response. 
  

• Review AI privacy policies: Familiarise yourself with the privacy policies of AI tools to understand how your data will be used, stored, and shared. Opt-out of any data-sharing practices that may compromise your privacy. 
  

 

Training and Awareness 

Clients should also invest in educating themselves and their teams about the proper use of AI tools. This includes: 

• Training programs: Organise training sessions to educate staff on the risks and best practices of using AI tools in legal processes. 
  

• Creating guidelines: Develop clear guidelines and policies for the use of AI tools, ensuring everyone understands the limitations and necessary precautions. 
  

• Regular audits: Conduct regular audits to ensure compliance with data protection, copyright, and privacy practices. 
  

 

Use Trusted AI Tools 

While ChatGPT is a popular choice, clients should also consider using AI tools that are specifically designed for legal purposes. These tools often have more robust mechanisms for ensuring accuracy and compliance with legal standards. Some steps include: 

• Researching available tools: Look for AI tools that are tailored to the legal industry and have strong reviews for reliability and accuracy. 
  

• Assessing vendor policies: Ensure that the AI tool providers have stringent data protection and privacy policies in place. 
  

• Seeking recommendations: Consult with legal professionals or industry experts to identify trustworthy AI tools that are appropriate for legal document review. 
  

 

Conclusion 

ChatGPT and other AI tools may seem convenient for legal document reviews, but they also pose serious risks. Clients must be cautious, understanding the limitations and potential legal pitfalls of using AI to review legal documents.  

Always verify AI-generated information independently, protect sensitive data, and comply with relevant privacy and copyright laws. By taking these precautions, clients can better navigate the complexities of using AI tools in their legal processes.