This section refers to what data is being collected by your interactions with the AI app and how that data is being used.
In general, most apps collect some personal information when you sign up for an account, minimally your email address and name. Computer information can also be collected including the operating system, browser, and IP address of your device. AI apps will also collect the chat prompts you give it and any files you upload into the app.
Privacy
This section refers to what data is being collected by your interactions with the AI app and how that data is being used.
In general, most apps collect some personal information when you sign up for an account, minimally your email address and name. Computer information can also be collected including the operating system, browser, and IP address of your device. AI apps will also collect the chat prompts you give it and any files you upload into the app.
Companies may use this information to train their AI models to improve their services. For this reason, ensure that any files or text prompts you give the app do not include any personally identifiable information (PII) including such things like social security numbers, college IDs, and dates of birth.
To determine what data is being collected and how it is being used, you can check the company’s privacy policy and/or check the settings within the app. There may be options to opt-out of training their models, whether to save your AI activity history, whether to share your data with third parties, and whether you’d like to delete your activity history.
Privacy best practices summary:
- Check to see what data is being collected and how it is being used.
- Check the privacy policy and/or tool settings for this information and for any opt-out options.
- Do not upload files with personally identifiable information (PII).
Ethics
This section will review copyright, attribution, bias, and influence.
For content that is generated by AI, there have been questions of whether it can be copyrighted. In Aug. 2023, the U.S. Copyright Office said AI lacks human authorship and cannot be copyrighted.
While AI-generated content cannot be copyrighted, it is still best practice to give attribution to it. Attributing provides transparency on where the information came from and gives the reader the right to gauge credibility. When citing, treat it as the output of an algorithm.
AI-generated content is produced based on datasets it has been trained on to deduce patterns and make predictions. Sometimes the generated content may not be entirely inclusive or may try to steer you in one way of thinking versus another. As you look at the content, ask yourself what you see, what you don't see, and is it truly representative of what you want to convey.
Ethics best practices summary:
- You cannot copyright AI-generated content, but you should still give it attribution.
- Review the output for any bias or influence: what do you see, what don’t you see, is this truly representative of what you want to convey?
Data Accuracy
This section includes information about cutoff dates and hallucinations.
Some apps may have a cutoff date. This is the last date an AI’s data set was last updated/trained. This is important to know when doing research as the generated output may not be inclusive of the latest information.
Sometimes the information generated from an AI tool is not accurate. When this happens, we say the AI app had a hallucination. Some tools will now include a link to the site where its information came from. Be cautious of tools that don’t include this.
Data accuracy best practices summary:
- Determine if the AI app has a cutoff date.
- Some fact-checking may be required due to hallucinations
Security
Just like AI is helping us become more productive, it too can help hackers be more productive. This section includes information about deepfakes and phishing.
Deepfakes is AI-generated content that imitates someone else, often for harmful purposes. Examples include robocalls using someone else’s voice or a zoom call where the participant looks and sounds like someone else.
AI also helps hackers create phishing emails. Harvard Business Review (HBR) conducted an experiment where they sent three different phishing emails: one that was generated by AI, one that was generated by a human expert in phishing emails, and one that was generated by AI and then edited by a human expert. The AI one had a 37% click-through rate, the human expert one had a 74% rate, and the AI-generated one then edited by a human had a 62% rate. HBR then predicted that the quantity of phishes would increase, the detection rate would be harder because spelling mistakes and bad grammar often used in detection would not be there anymore, and the click through rate will increase.
Security best practices summary:
- Be suspicious. If something doesn’t seem right, it probably isn’t.
- Review Campus Technology Services’ (CTS) phishing identification tips
- Call the CTS Help Desk for assistance.
Sustainability
Sustainability concerns include energy consumption and emissions. AI’s energy consumption and environmental impact are still being researched. One researcher states that one ChatGPT query takes up the same amount of energy as one light-bulb that stays on for about 20 minutes. Another researcher says that there will be an impact, but might not be as much as we think. Companies are working to reduce their carbon footprint as much as possible while still keeping up with growing demand for AI.
Sustainability best practices summary:
- Determine if a regular search vs. AI is applicable. However, even searches now include results generated by AI.
- Hope innovations will take place to reduce energy use and companies reach their goals of reducing emissions.