AI compliance statement

One part of SideKrew School uses AI: the hint-first AI helper, and the test checks and practice questions that go with it. They run on Claude, made by Anthropic. Everything else (the list of work, plans and reminders) works without AI.

AI help is on only for students on a pilot school’s roster, in classes whose teacher said yes in writing, and, for every student under 18, only after a parent says yes.

• Canada’s Personal Information Protection and Electronic Documents Act (PIPEDA).
• Ontario’s Child, Youth and Family Services Act (CYFSA), including the duty to report a child who may need protection.
• Canada’s Anti-Spam Legislation (CASL), for email.

SideKrew is a private company, so Ontario’s Municipal Freedom of Information and Protection of Privacy Act (MFIPPA) doesn’t bind us directly. It binds school boards, and a board’s own duties run through the board.

• Anthropic’s Usage Policy, including its requirements for organizations whose products serve minors.
• Google’s API Services User Data Policy, including the Limited Use requirements.

• The AI helper’s screen always shows “AI helper · not a person”.
• Every session starts by saying so, and says it again every 3 hours.
• The student’s start page and the parent’s consent form say what goes to Anthropic, in the United States.
• Every session ends with a note the student can copy to credit the help.

• It gives hints, and worked examples on similar problems, never answers to assigned questions or finished writing. A check on every reply blocks complete answers.
• Quizzes, tests and Classroom questions get topic practice, not help with the item.
• Every message is screened for safety before anything else runs. High-risk messages never reach the AI: the student sees crisis lines, and a person on our safety team reads the message.
• If the AI raises a safety concern the screen missed, its reply is thrown away and the same safety steps run.
• Classmates’ names are filtered out before anything goes to the AI, and the student’s name and email are never sent.
• Instructions hidden in class posts are treated as text, not as commands.
• A daily message limit and a daily cost cap per student. Safety messages are never blocked or counted by a limit.
• Report this chat is one tap away in every session. A person on our safety team reads every reported chat.
• Release tests on teaching quality, safety and hidden instructions must pass before AI help is turned on.
• If Anthropic provides a child-safety system prompt, we add it.

AI help is open only while a person on our safety team is on call, during the hours your school set, Toronto time. Crisis lines show at every hour.

We never report a student’s help requests to a school or a parent as cheating. Parents never see help chats; they see only how many help sessions were used.

Requests are processed by Anthropic in the United States, and Anthropic doesn’t use them to train its models. How long Anthropic keeps them, and everything else we do with data, is in our privacy notice.

Questions about how we use AI: privacy@sidekrew.com.