DEMYSTIFYING: What puts the intelligence in Artificial Intelligence
AI has become the shorthand for any complicated automated process, but there are important distinctions. Traditional automated systems, such as your typical law practice management software, operate on fixed, predefined pathways. At the core of AI tools (e.g. ChatGPT) is machine learning, which improves responses over time, mimicking the way humans learn from experience. It is important to understand that it does this by analyzing vast amounts of data and making adjustments using mathematical models to better predict outcomes. This process allows AI to handle new, unexpected scenarios with a degree of finesse that complex automation cannot match. However, it's crucial to note that AI's learning is not mimicking human cognitive processes in a literal sense in that its adjustments are based on mathematical models.
RISK: What AI is learning from you, it may be sharing with others
There is a real concern that confidential data – about cases or clients or proprietary info – that you feed into an AI system could be leaked or shared with other users. (Look no further than the reported impetus behind Samsung’s ban of ChatGPT for an example of this.)
But you can mitigate that risk!
MITIGATION
Deploy safeguards to prevent sensitive, privileged or confidential data from being retained and shared with other customers:
- Limit use to the many AI tools which offer subscriptions to their products that allow you to prevent the tool from learning based on your data (e.g., Microsoft Copilot for Business or ChatGPT Enterprise).
- Anonymize or remove personal information when using AI tools that do not permit disabling training.
Additional reading: To Use or Not to Use: Navigating privacy risks associated with generative AI tools (James G. Kosa, Vipal Jain and Yatin Sidhu, Weirfoulds LLP, JUST magazine, March 2024)
POLICY CHECKLIST FOR YOUR ORGANIZATION
- Make it required practice to review the provider’s privacy policies and terms and conditions before implementing any AI tool in your practice to ensure you are comfortable with the way your data will be used and stored.
- If you have an Enterprise version, make sure your staff only use the Enterprise version.
- If you’re not using an Enterprise version, never put information in that will reveal directly or indirectly the name of your client or any other confidential or privileged information.
- Remind staff to consider the aspects of the transaction or case that are unique and that will identify your client or sensitive information.
-
did you know…?
When it comes to generative AI, ChatGPT remains among the most ubiquitous tool. You can get started with ChatGPT by visiting its introductory page on the Open AI site and trying out a (non-confidential) question or two for fun. There are four versions available (including, ChatGPT Free, ChatGPT Plus, ChatGPT Team and ChatGPT Enterprise). If you’re looking to deploy ChatGPT in your legal practice, you should consider subscribing to a paid version like ChatGPT Team or Enterprise, which offers access to Chat-GPT 4 (their newest release), the ability to create a shared workspace, along with privacy controls that allow you to exclude your Team data from “training” by default.
COME TO OUR SERIES
HAVE YOU TRIED THIS FOR FUN…?
Using AI to craft compelling invitations for a must-attend event? (See Lenczner Slaght’s HolidAI Card generator – aka Summons Bot – for a creative example, litigator style.)
Look for more Real Intelligence tomorrow!
Have your own tip you would like to share? Email it to us for possible inclusion in an upcoming issue of our Real Intelligence on AI newsletter. You can continue to follow tips, developments, perspectives and best practices in our regular Real Intelligence on AI newsletter, JUST. magazine, and CPD courses focused on AI.