Futureproof Festival would be a ridiculous place to pretend the machines are not here. We use them to research, draft, design, code, test, and answer festival questions. We do not hand them the steering wheel, the guest list, or the final word.
Use the machines. Keep your name on the decisions.
AI is allowed to be fast, strange, useful, and occasionally brilliant. It is not allowed to become an excuse for fake certainty, borrowed personality, invisible labour, sloppy consent, or nobody being accountable when something lands badly.
The exact vendors will change. The responsibilities do not. These are the current lanes where AI materially helps us make and operate Futureproof.
We use AI to organize research, test arguments, find gaps, tighten drafts, and get unstuck. The brief, point of view, sources, and final responsibility stay with a person. We do not feed someone else’s work into a model and call the rearranged version ours.
AI agents help us write code, inspect systems, reconcile records, and automate repetitive work. Their changes go through the same tests, review, and privacy rules as anything a person writes. Fast is useful. Untraceable is not.
We use generative image tools as part of Futureproof’s campaign-art practice. We do not pass synthetic scenes off as documentary photographs, fabricate attendance or participation, or invent a person’s face, quote, endorsement, or presence. Permissioned likenesses and final public assets get human review.
When enabled, FestivalChat uses Anthropic’s API to answer questions from a bounded set of public festival facts. It cannot see registrations, payments, private attendee data, unannounced speakers, or your account. It can still be wrong. Sensitive questions go to a person.
Names, dates, prices, quotes, credits, partnerships, availability, and claims get checked against a real source. Fluency is not evidence. When we get something wrong, we correct it.
We do not clone a voice, fabricate a likeness, manufacture a testimonial, or place someone at the festival without permission. Stylized speaker art begins with permissioned source material and stays private until a human clears the exact asset.
We minimize what a tool receives and do not paste sensitive personal, payment, health, accessibility, credential, contract, or protected roster information into public consumer AI tools. Anything public-facing only sees what we have already published.
AI does not make the final call on speakers, press accreditation, sponsorship, prices, refunds, accessibility, consent, safety reports, conduct enforcement, publication, or external messages. People own decisions that affect people.
If you are interacting with an AI system on a Futureproof surface, we say so. We explain its lane, keep a human contact available, and test the boundaries instead of relying on a cheerful disclaimer.
When the chat is enabled, the site labels it as an AI assistant. It only knows what is already published on this site. We test it hard for leaks, invented claims, and prompt games, and it still gets things wrong sometimes.
Your message and the recent conversation are sent to Anthropic's commercial API to generate an answer. Futureproof's site does not write chat transcripts to its own database. Do not put payment, health, accessibility, confidential, or other sensitive information in the chat.
Anthropic says commercial API inputs and outputs are not used for model training by default and are normally deleted from its backend within 30 days, with published exceptions for abuse monitoring, legal obligations, feedback, and different contractual settings.
Lara Kroeker's policy gave us the shove to write this plainly. The other sources helped pressure-test the parts about accountability, privacy, verification, disclosure, creator rights, and ongoing evaluation. We borrowed the questions, not their sentences or institutional posture.
The spark: a plainspoken, first-person policy that names the real tools and the real line between looking, learning, and taking.
Accountability, security, transparency, education, privacy, source verification, and restraint when the risk cannot be managed.
AI can assist with bounded tasks; editorial judgment, verification, and accountability remain human responsibilities.
Use AI for a specific benefit, keep human oversight, disclose material use, and respect the people who create and own source material.
Define human roles, test systems, manage privacy and information-integrity risks, and keep monitoring after launch.
Good. This is a living operating policy, not a marble tablet. If something here is unclear, untrue, or not good enough, email hello@bc-ai.ca. We would rather answer a sharp question than hide behind a soft paragraph.
