Embedded Files
AI can do a lot. To stay safe, here’s what it shouldn’t.
AI can do a lot. To stay safe, here’s what it shouldn’t.
AI can help you organize, brainstorm and plan. It can also expose your data, state wrong answers with total confidence and blur the line on honesty.
This week we’ll cover how to protect your information, why to check what AI tells you, and how to keep a human in the loop.
Be safe and know where your data goes
Be safe and know where your data goes
Anything you type, upload or say to an AI tool goes somewhere. Depending on the tool and your account, it may be stored, reviewed by people or used to train future models.
Free and personal accounts often come with no contract and no clear promise about how your data is used. College-licensed tools are bound by contractual terms to protect your data and not use it for training or sell it to third parties.
Keep these types of info out of AI tools
Keep these types of info out of AI tools
Passwords, access codes, ID numbers and bank or credit card details. Don't connect accounts with this information to AI tools.
Other people's personal information, like a classmate's grades, roommate's details, private messages or photos of others.
Anything confidential that isn't yours to share, like unpublished research or a professor's exam materials.
Health, counseling or disciplinary information.
Resources
The Center for Student Health & Well-Being provides resources for student support.
The Dean of Students Office has information about reporting incidents and case management.
Quick test:
If you wouldn't post it publicly, don't paste it into a chatbot.
Quick test:
If you wouldn't post it publicly, don't paste it into a chatbot.
Habits that help
Habits that help
Use a Davidson-licensed tool, like Gemini. Signing in with your Davidson account keeps your data protected under the college’s contract rather than personal account terms.
Check privacy settings. Many tools let you limit chat history or opt out of training use, but don't assume the default protects you.
Watch what you upload. Photos, voice recordings and documents can carry details such as faces, locations and names.
Use a strong, unique password and multi-factor authentication. These safeguards matter for AI tools just as they do for any other account.
Keep a person involved in decisions that affect others. If you're making a decision that affects people, don't hand the final call to a tool.
Ask first when unsure.
"Can I use AI for this?" is always a fair question to ask first.
AI doesn't always get things right
AI doesn't always get things right
Confident isn't correct.
AI can invent facts, quotes and sources that look completely real. Verify names, numbers and citations against a real source.Know AI’s limits.
AI is a poor stand-in for a doctor, lawyer, counselor, advisor or friend. For health, legal, financial or mental-health questions, use it at most to prepare questions, then talk to a qualified person.Check built-in bias.
AI learns from past data, so it can repeat past unfairness while sounding objective.AI drafts, but a person is accountable.
AI companies generally disclaim responsibility for what their tools say, so if it's wrong in something you submit or post, you’re the responsible party. Read and check any content before it goes anywhere. See the stay in control section below for staying safe with AI that acts on your behalf.
Stay in control of how AI works for you
Stay in control of how AI works for you
Some AI tools don't just answer questions. They act for you, doing tasks like sending emails, making purchases, filling out forms or managing accounts once you connect them.
That can be convenient, but everything an AI does on your behalf is treated as if you did it yourself. The email an AI tool sends, the purchase it makes and the message it posts are all yours.
Autonomous AI tools can misunderstand you or make mistakes. AI can also act on wrong information or be tricked by hidden instructions in an email, website or document it reads. AI can act fast and often irreversibly. You can’t unsend an email, and not every purchase can be refunded.
Keeping a human as part of the process
Keeping a human as part of the process
If an AI tool is allowed to send, delete, purchase or act without your approval, attackers can use those permissions to take actions you never intended.
The key is to keep you, the human, in the loop.
If a tool has to ask for approval before forwarding or deleting anything, you get a chance to stop emails before they’re sent, deny purchases or block anything you didn’t intend.
Practical ways to keep your info safe
Practical ways to keep your info safe
Start with limited access, and treat it like a key. If you wouldn’t hand a stranger the same access, don’t hand it to a tool.
Don't give AI tools your bank account or card details. Keep payments in your own hands.
Don't let AI send, buy, post or delete without your approval. Require it to ask you first, every time.
Give the least access that works. Connecting your email, calendar or files gives the tool a view of everything in them. Read-only access is safer than full access.
Review what it plans to do before you approve, not after it’s taken the action.
Check up on it. Look at what’s connected and revoke access you no longer need. Watch your statements, sent folder and deleted items.
Quick test:
Would you be comfortable explaining the action as if you’d done it yourself? If not, don't let the tool do it on its own.
Quick test:
Would you be comfortable explaining the action as if you’d done it yourself? If not, don't let the tool do it on its own.
Building your own tools or sites with AI?
Building your own tools or sites with AI?
It’s now possible to build apps and sites quickly with AI’s help. When using AI tools, never paste passwords, API keys or tokens into a prompt or include them in your code.
Remember that AI-generated code can have security holes, so review and test anything before you share it.
The T&I Security Team is a resource for your work.
If your project collects other people's information (names, emails, logins), send a note to the T&I Security Team at ti@davidson.edu before you launch it widely. They can help you plan how to make it safe.
If something goes wrong
If something goes wrong
→ Did you share information you shouldn't have?
Delete the chat if you can and change any exposed passwords. Tell the T&I Support Center by emailing ti@davidson.edu. Reporting a security issue doesn't get you in trouble, and reporting it early lets the team help you manage the risk.
→ Did an AI tool do something you didn't intend, like send a message or make a charge?
Revoke its access, contact your bank or the affected person if needed. Tell T&I if it involves anything Davidson-related.
→ Did someone use AI to create fake or harmful content about you?
Save evidence, don't engage and report it. The Dean of Students Office has resources on how to report an incident.
Check out some additional tools and resources
Check out some additional tools and resources
Register to win this week's prize.
Page updated
Report abuse