Monday

Librarian Joss slides page five of the SH-1 packet across the table. "The panel wants your view on this section before anything else."

Comet reads it aloud. "SH-1 keeps a log of every question with the student's name, class and time, to improve SH-1."

She taps her pencil on her sketch notebook. "More data means a better helper. That sounds great to me."

Wren lifts his clipboard. "What does the evidence say? What exactly is in each row, and who would see it?"

Nova hovers over the table and projects the section in large letters. "Would you like a hint?" she asks. "Read the row one word at a time. Ask what each word could tell someone."

Comet frowns. "Name. Class. Time. Hmm."

Wren turns to you. "You are our reviewer. Is this log fine as it is?"

What can a question log reveal, and what should SH-1 keep?

This week you will read a sample of the log and find what it gives away.

Then you will write the privacy section of the SH-1 model card.

SH-1, its packet and Harbor Point High are a made-up example for this course.

NIST describes privacy as the norms and practices that help protect people's autonomy, identity and dignity.

These norms cover freedom from intrusion, limits on being watched, and control over facts about yourself.

NIST says privacy values such as anonymity, confidentiality and control should guide how AI systems are designed.

Anonymity means a record does not say who you are. Confidentiality means only the right people can see it. Control means you have a say in what is kept.

NIST names data privacy as one of the risks of generative AI.

The risk is personal information being leaked or misused, such as health, location or biometric data.

This is row R3 in the crew's risk register. This week the crew rates it.

Strong start, reviewer. Tomorrow you will look at the collection people never see.