ChatGPT and Your Privacy
The latest in artificial intelligence (AI) has arrived in the form of ChatGPT. This has spawned new interest in the use of AI to create and maybe even replace human workers or entire companies. The results when testing the platform are fantastic; I can ask it to write me a song, write a poem, tell me who the best basketball player is (Michael Jordan), and predict the outcome of March Madness.
But, how does the application know so much? How is it able to respond so quickly with seemingly correct or near correct responses?
The answer lies in the sheer amount of information which ChatGPT is fed. The application is full of information scraped, collected, and found unprotected on the Internet. Any blog posts, reviews, responses, articles, or other posts on the web unprotected are most likely part of the underlying repository of information being accessed by or fed to ChatGPT.
Asking ChatGPT the question: “how do you know what you know,” resulted in the following response:
As an artificial intelligence language model, I “know” things based on the data I was trained on. Specifically, I was trained on large datasets of text from various sources such as books, articles, websites, and other documents. The training process involved processing and analyzing the text to recognize patterns and relationships between words and phrases, which allows me to understand and generate text.
All the information gathered and fed to ChatGPT becomes part of datasets, its knowledge, from which it draws its conclusions. But what if the datasets include information which is copyrighted, trademarked, or personal information that was previously deleted from a blog or site?
As privacy laws are being adopted or being considered by states, such statutes usually include the right to be removed from a company’s database. Take for example the California Consumer Privacy Act as amended by the California Privacy Rights Act, which provides that “[c]onsumers should have access to their personal information and should be able to correct it, delete it, and take it with them from one business to another.” If personal information is incorporated into a ChatGPT dataset how can a consumer request that their information be deleted?
Let’s ask the expert: “ChatGPT, how can I delete personal information you may have in your dataset which is identifiable with me?”
As an AI language model, I don’t have access to personal information of individuals, and I don’t store any data on my own. However, if you believe that your personal information has been collected by a company or organization and you want to have it deleted, you should follow these steps:
- Identify the company or organization: Determine which company or organization has collected your personal information.
- Contact the company: Reach out to the company via email or phone and request that your personal information be deleted. Be sure to include your name and any other information that can help them identify your data.
- Follow up: If you do not receive a response from the company, or if they do not fulfill your request, you may need to follow up with them. You may also consider submitting a complaint to the appropriate regulatory agency.
- Monitor your personal information: Keep an eye on your personal information to ensure that it is no longer being collected or used by the company. If you find that your personal information is still being collected or used, you may need to take additional steps to protect your privacy.
This answer really amounts to the fact that information cannot be deleted. Once the personal information becomes incorporated into the underlying datasets, the information is virtually immutable and unchangeable. This permanence of such information in a dataset can become a nightmare for companies wishing to use or incorporate ChatGPT or any AI model into its business.
There are many legal issues implicated in using ChatGPT that need to be addressed before these technologies are widely adopted. These issues include:
- How will AI deal with or identify inaccurate information?
- Will the methods for collecting and ingesting information be documented and reviewed (and ultimately regulated)?
- How will AI manage trademarked or copyrighted material?
- Does the data scraping from websites violate any contractual requirements of these companies?
Many questions need to be answered before businesses adopt AI for their day-to-day use, but until then: “ChatGPT, who will win March Madness this year?”