One of the most potent capabilities of generative AI is its capacity to assimilate human queries, making it a potential minefield for attorneys striving to leverage the technology while safeguarding confidential and privileged client data.
Attorney-client privilege safeguards specific confidential exchanges between legal counsel and clients seeking legal guidance. However, this privilege can be forfeited if the information is shared with external parties, potentially leaving the communication vulnerable to discovery by opposing parties during legal proceedings.
The use of public-facing generative AI models, such as ChatGPT’s free version 3.5, presents a concrete risk to confidential information. These models have the potential to regurgitate information from one user’s query to another user inquiring about similar topics.
Steve Delchin, a senior attorney at Squire Patton Boggs, highlighted the concern by stating, “The problem is, the next person that comes along might be your opposing counsel,” suggesting that skilled adversaries could extract information from the AI model.
Delchin emphasized that confidentiality is a paramount obligation when lawyers engage with generative AI technologies. The mere possibility of third-party exposure to sensitive data by using public-facing tools could constitute a breach of legal ethics, akin to leaving classified documents unattended on a public bench.
Chief Justice John Roberts also underscored these concerns in his annual report on the federal judiciary, acknowledging the apprehensions raised by legal scholars regarding the potential compromise of legal privileges when inputting confidential information into AI tools.
Enterprise Solutions and Data Segregation
AI developers have acknowledged the necessity for businesses, including law firms, to segregate their data. They offer enterprise models that exclusively train on a company’s data, ensuring that the information remains within the firm’s confines without feeding back into public models. However, this approach does not entirely eliminate the risks associated with confidentiality and privilege.
For instance, if a law firm deploys a generative AI model trained solely on internal data and queries, there remains a concern that accessing this information within the firm could inadvertently waive attorney-client privilege, as noted by Nick Peterson of Wiley.
James McPhillips, a partner at Clifford Chance’s global technology group, suggested that using an internal AI model to draft public-facing documents might not pose a problem. However, he raised a valid query about the potential crossover of information between clients when generating specific legal clauses.
To address these challenges, AI platform developers are evolving their solutions to allow for greater customization and data segregation within organizations. Megan Ma from the Stanford Program in Law, Science, and Technology highlighted the trend towards building “smaller walls” around data to mitigate risks and enhance privacy measures.
Considerations for Client-Facing Chatbots
Law firms implementing client-facing chatbots on their websites need to exercise caution to preserve confidentiality and privilege. Ken Withers from the Sedona Conference cautioned that if data entered into a chatbot is not isolated and flows back into a public model, it could jeopardize confidentiality.
Withers also raised a thought-provoking question about the chatbot’s role and sentience in client interactions. He pondered whether the chatbot functions merely as a communication tool or potentially as a conversational participant, which could impact the application of attorney-client privilege.
Due Diligence with Legal Tech Vendors
As legal tech vendors integrate generative AI capabilities into their products, attorneys must scrutinize the controls and restrictions available to safeguard data segregation. Ron Hedges, a former US magistrate judge, stressed the importance of understanding how vendors handle confidential information and what measures are in place to preserve confidentiality and privacy.
Attorneys should inquire about data storage, access protocols, ownership rights, security safeguards, and data usage policies before engaging with AI vendors. The evolving landscape of AI in the legal profession may lead to anticipated litigations to clarify the implications of generative AI tools on legal practice.
In conclusion, while concerns persist regarding the confidentiality and privilege risks associated with generative AI, informed use of these technologies, coupled with stringent data protection measures, can help mitigate potential legal and ethical breaches.
