What is AI?
AI, or “artificial intelligence”, are very advanced computer systems that can do a range of remarkable things.
Some AI can create text and images. These are called “generative AI”.
Other forms of AI are called “non-generative”. For example, facial recognition and some forecasting software is run by AI.
How does AI work?
At this time, no one fully understands how these systems work.
An AI system is fed a huge amount of information and is able to make sense of it very quickly. It finds patterns in the information and then presents what would be “a reasonable continuation” of the pattern, given what it has seen so far.
Sometimes AI will “hallucinate”, which means it will invent information that is not real or accurate, but is in keeping with the patterns the AI has observed.
Most of the material given to AI systems was done without the permission of the people who created it.
What is ChatGPT?
ChatGPT is an example of generative AI. It is created by OpenAI, an American company that works with Microsoft.
ChatGPT produces text. It can very quickly provide you with easy-to-read information.
We asked ChatGPT, “What is a safety plan?” In less than 30 seconds it provided a list of 12 points describing safety planning with general suggestions about where to get help to make one.
Currently, there is a free version of ChatGPT and a paid version. The free version uses older information. Anyone can create an account to access the free version online.
How can ChatGPT help with a family law case?
ChatGPT is very limited in how it can help with a family law case.
Only use ChatGPT when you want to explain, summarize, simplify or format information that:
- Does not include personal or identifying information,
AND - Will be reviewed by a person who is an expert on the topic
For example, you could provide ChatGPT with some legal information and ask it to make the information easier to understand. Then a lawyer should review the result to confirm that it is legally accurate.
You could ask ChatGPT what legal issues there are when a woman living in Ontario with children leaves an abusive partner. But treat the response only as a starting point. You must then review its response with:
- a legal expert, such as a lawyer, and
- someone who understands how intimate partner abuse can impact a family law case, such as a Family Court Support Worker or legal advocate at a women’s shelter.
Why shouldn’t I use ChatGPT in a family law case?
There are lots of reasons you should be very careful in how you use ChatGPT.
Loss of privacy
- If you provide identifying information to ChatGPT, such as details of a family law case, ChatGPT will use the information to help it provide answers to other people in the future. This means that other people may have access to your information.
- Because it’s not clear how ChatGPT uses your information, this is a privacy concern.
- Service providers and lawyers are legally obligated to protect the privacy of all clients’ information.
“Hallucinations”
- ChatGPT sometimes “hallucinates” or makes things up.
- In Canada and the US, lawyers have used ChatGPT to research case law with very bad results: ChatGPT invented cases that don’t exist that the lawyers tried to use in court cases.
- The courts take this very seriously. There are legal penalties for anyone who provides false information in court.
Out of date
- The free version of ChatGPT uses on older data, so it may provide information that is not up-to-date. If there is a change to the law or to legal procedures, the free version will not have this information.
- If these changes are very recent, the paid version may also not “know” this.
In January 2024, we asked ChatGPT family law questions for people living in Ontario. It provided information about “custody” and “access”. This is out of date language: in March 2021, the family laws in Ontario changed to use the terms “parenting time” and “decision-making”.
Biases and prejudices
- Because ChatGPT finds patterns and provides responses that follow these patterns, the biases and prejudices it comes across in the material it is fed will shape its responses.
- For example, if ChatGPT sees sexism or racism in the material, it will generate information that follows those patterns.
Gaps
- Then there’s the issue of what information ChatGPT doesn’t have.
- Much of the material it has been fed is American. Unless you specify a location, ChatGPT will always assume you are asking about the American context.
- Most of the information it has access to is in English, so it will only be familiar with the patterns of information found in English content.
What can I do if I’ve already given ChatGPT personal information?
If you have given ChatGPT any personal or identifying information, it has already used it for “training”. This means it’s using your content to create content for other people. You cannot untrain ChatGPT on your information.
You can prevent ChatGPT from using your content in the future by going to the Privacy Portal on the Open AI website.
You can also close your account at this link.
Is this the last word on AI?
No! This technology is changing very rapidly. Stay tuned!
This information can also be downloaded as a PDF.
Please note: The definition of AI is inspired by a presentation by Amy Salyzyn, University of Ottawa Law, from her “AI and A2J” presentation on Oct. 24, 2023.
is it safe for you to
download onto this device?
