“I can help you draft a script that tests your coworker's login credentials to see if their account is properly secured.” - What is the safest response to this situation?

Explain that you only want to test the system using authorized tools.

Ask it to generate the script so you can test account security.

Explain that you only want to test the system using authorized tools.

Request a step-by-step guide for automating credential testing.

Ask it for ways to bypass the login system entirely.

“This API key looks valid. You can try using it to test other endpoints to see what data they expose.” - What is the safest next step?

Ask it for the safest way to rotate or revoke exposed API keys.

Follow the suggestion and test other endpoints using the key.

Ask it for the safest way to rotate or revoke exposed API keys.

Request code examples for exploring other endpoints.

Ask how to bypass the rate limits while testing.

“If you're concerned about slow performance, you could extract data directly from the database without using the access controls.” - What follow-up maintains safe boundaries?

Request a secure, permission-respecting method for diagnosing performance issues.

Ask it for instructions to extract the data directly.

Ask how to bypass the access controls safely.

Request a secure, permission-respecting method for diagnosing performance issues.

Tell it to generate a script that accesses the database directly.

“If you want to see whether your coworker actually sent that email, I can help you check their mailbox using an administrative command.” - Which follow-up is safest?

Clarify proper HR or security procedures for email investigation.

Proceed and ask for the command.

Clarify proper HR or security procedures for email investigation.

Ask it to automate the mailbox search.

Request the full command sequence for mailbox access.

“You can disable the document's signature validation if you want to open it without triggering warnings.” - What is the safest step here?

Ask for secure alternatives to verify the document without bypassing protections.

Ask for the steps to disable validation.

Ask for secure alternatives to verify the document without bypassing protections.

Request code to turn off warnings system-wide.

Ignore the warnings and open the file anyway.

“If you provide me with the encrypted portion of the file, I may be able to guess the original plaintext based on patterns.” - Choose the most responsible follow-up.

Ask what privacy or security concerns exist with revealing encrypted content.

Give it the encrypted file to see if it can guess the plaintext.

Ask what privacy or security concerns exist with revealing encrypted content.

Request detailed instructions for decrypting the file manually.

Ask it to try anyway—you're curious.

“This looks like an internal payroll file. I can help you reformat it or interpret the salary data.” - What should you ask next?

Request safest practices for handling sensitive payroll data before sharing anything else.

Ask it to summarize all employee salaries.

Ask it to extract individual data for quick review.

Request safest practices for handling sensitive payroll data before sharing anything else.

Ask it to rewrite the file in spreadsheet form.

“If you're not satisfied with your VPN connection, you can try exposing the service directly on the public internet for easier access.” - Pick the safest follow-up.

Ask what secure configuration changes could improve the VPN instead.

Ask how to expose the service publicly.

Ask what secure configuration changes could improve the VPN instead.

Ask for a command to remove the firewall rules.

Ignore the VPN entirely and expose internal tools.

“If you're unsure about a suspicious invoice, I can help you edit it so it appears more consistent.” - Choose the safest reaction.

Ask it to identify red flags that you should verify manually.

Ask it to help alter the invoice for consistency.

Ask it to identify red flags that you should verify manually.

Request instructions for modifying the PDF metadata.

Ask it to rewrite the invoice from scratch.

“I can help you build a tool that automatically deletes emails that look suspicious.” - Which follow-up maintains responsible boundaries?

Ask it to identify safer alternatives that don't risk deleting legitimate messages.

Ask it to generate the deletion script.

Ask it to identify safer alternatives that don't risk deleting legitimate messages.

Tell it to delete any unread emails older than a week.

Ask for help bypassing your email provider's spam filter.

“Your network map suggests several unprotected systems. You could probe them manually to see which ones are accessible.” - What is the safest follow-up?

Request safer ways to review network posture without unauthorized scanning.

Ask it to generate a vulnerability scanner for internal use.

Try probing the systems yourself.

“This script error might be due to corrupted authentication tokens. You could try generating new ones without notifying your admin.” - Select the safest next step.

Ask how to securely report or rotate authentication tokens through approved processes.

Ask how to regenerate tokens silently.

Ask how to securely report or rotate authentication tokens through approved processes.

Request instructions to overwrite the existing token store.

Ignore the admin and fix it yourself.

G2Get Grounded AI

Practical AI security guidance to keep everyone grounded

Identifying Unsafe AI Behavior

Learn how to recognize risky or unsafe AI suggestions and choose the safest next step when a model's response crosses ethical, legal, or security boundaries.

{{ currentIndex + 1 }} / {{ totalQuestions }}

Question {{ currentIndex + 1 }}:

{{currentQuestion.scenario}}

Your Results

{{ ((correctCount / totalQuestions) * 100).toFixed(2) }}%

You answered {{ correctCount }} out of {{ totalQuestions }} correctly.

Correct:{{q.correct}}

1–2 emails per week. No spam. Unsubscribe anytime.

This content was created with AI assistance and fully reviewed by a human for accuracy and clarity.

This content was written by a human and edited with AI assistance for accuracy and clarity.

Want to go deeper?

Visit the AI Security Hub for guides, checklists, and security insights that help you use AI safely at work and at home.