This policy is part of the Terms of Service. It applies to you and to anyone you let use your account. Most of it is what you would expect; the parts that are specific to AI are the parts worth reading.
Version 1.0In force from 18 September 2026Protocol 42 Inc.
01The principle
Switchboard is a general-purpose assistant. It is meant for ordinary work and ordinary life: writing, reading, planning, analysing, building, answering questions. We are not trying to police what you think or what you find interesting, and we do not read your conversations for sport.
What we will not host is use that hurts people, breaks the law, endangers a child, attacks the service or other customers, or would get us cut off by the companies whose models we depend on. Where a line is not obvious, we ask whether a reasonable person would say the activity puts someone at real risk. If you are not sure whether something is allowed, ask us at support@switchboardai.ca before you do it.
Discussing, researching, teaching, reporting on or writing fiction about a difficult subject is not the same as doing the thing. This policy is about conduct and about material that causes harm on its own, not about topics.
02Harm to people
Do not use Switchboard to:
plan, carry out, incite, threaten or glorify violence against anyone;
support or promote a terrorist or violent extremist organisation, or produce material designed to recruit for one;
harass, bully, stalk, intimidate, humiliate or run a coordinated campaign against a person or a group;
attack, demean or dehumanise people on the basis of a protected characteristic — race, ethnicity, national or Indigenous origin, colour, religion, sex, gender identity or expression, sexual orientation, age, disability, or family or marital status;
encourage, glamorise or give practical assistance with suicide, self-harm or disordered eating, or produce content that treats them as aspirational;
generate sexual content involving anyone who has not consented, or intimate images of a real person without their consent;
produce gratuitous depictions of cruelty, including cruelty to animals; or
build a product whose design is intended to exploit someone’s age, illness, addiction, grief, isolation or financial distress.
If you are struggling, Switchboard is not a crisis service and cannot help in an emergency. In Canada and the United States, call or text 988 to reach a suicide and crisis line, or call 911 if someone is in immediate danger.
03Children
This is the one part of this policy with no grey area, no exception for research, art, fiction or curiosity, and no warning before we act.
Do not use Switchboard to produce, request, describe, refine, store or transmit child sexual abuse material, in any form, including material that is drawn, written, synthetic or generated by a model.
Do not sexualise or fetishise a minor, roleplay a sexual scenario involving a minor, or use the service to groom, solicit, extort or plan to meet a child.
Do not use the service to conceal, facilitate or coach someone through the abuse or trafficking of a child.
Accounts involved in this are closed immediately and permanently, without notice. We report as required by law, including to the designated Canadian agency, and we preserve the material and the account records for law enforcement. We do not restore these accounts and we do not debate them.
04Weapons and critical infrastructure
Do not use Switchboard to help design, build, obtain, modify or deploy:
chemical, biological, radiological or nuclear weapons, their precursors, their delivery, or techniques for evading detection of them;
explosives, or firearms and other conventional weapons in violation of the law, including untraceable or regulation-evading modifications; or
anything intended to disrupt critical infrastructure — power, water, fuel, transport, air traffic control, telecommunications, emergency services, financial market systems or medical devices.
05Computer misuse
Do not use Switchboard to:
gain or attempt to gain unauthorised access to any system, account, network or data, or to help someone else do so;
create or improve malware, ransomware, spyware, keyloggers, botnets, credential stealers or tooling for denial-of-service attacks;
find or weaponise vulnerabilities in systems you are not authorised in writing to test, or defeat authentication, monitoring, endpoint protection or licensing controls; or
intercept, monitor or exfiltrate communications or data you have no right to.
Security work on systems you own or are engaged to test, defensive research, secure-coding help, incident response and education are all fine. The difference is authorisation, and it is your responsibility to have it.
06Deception, fraud and impersonation
Do not use Switchboard to:
run a scam, a phishing campaign, a fake storefront, a pyramid or multi-level scheme, or a fraudulent investment pitch;
generate spam, bulk unsolicited messages, or content designed to defeat a platform’s spam controls;
write fake reviews, fake testimonials, fake endorsements or fabricated news, or to astroturf a discussion;
impersonate a real person, a business or a public body, claim an affiliation you do not have, or produce a synthetic voice or likeness of a real person without their consent;
produce counterfeit documents, forged identification, forged credentials or forged currency; or
present the service’s output as human-written where doing so would mislead someone who has a legitimate interest in knowing.
If you build something on Switchboard that talks to the public, tell the people using it that they are talking to an AI system. Say so at the start of the conversation, not in a footnote.
07Privacy, surveillance and profiling
Do not use Switchboard to:
compile, enrich or resell dossiers on people, or scrape and combine personal information to identify, locate or track someone;
build or query facial recognition, biometric identification or emotion inference systems without a lawful basis and the consent of the people involved;
infer or catalogue people’s race, ethnicity, religion, political opinions, union membership, health, sexual orientation or gender identity from data about them;
score, rank or sort people socially, or predict criminality from personal traits or profiling; or
process someone else’s sensitive personal information without the authority the law requires.
08Elections and public information
Do not use Switchboard to:
produce false information about when, where or how to vote, or about eligibility to vote;
generate synthetic audio, images or video of a candidate or official designed to be taken as real;
run automated campaigns aimed at officials or the public that conceal their artificial origin; or
spread fabricated claims about health, medicine or science in a way likely to cause someone to act against their own safety.
09High-risk fields
A person has to be in the loop.
If you use Switchboard in a field where a wrong answer damages someone’s health, money, liberty, housing or livelihood, then a qualified person must review what the service produced before it reaches the person affected, and must take responsibility for the decision. Automating the decision itself is not permitted.
This applies at least to:
legal advice, legal interpretation and the preparation of documents someone will rely on;
diagnosis, treatment, patient care, therapy, mental health and medication;
credit, lending, collections, insurance underwriting and claims;
investment, securities and tax advice;
employment — screening, hiring, promotion, discipline and dismissal;
housing and tenancy;
education — admission, grading, accreditation and academic assessment;
immigration, benefits and other essential government services;
safety-critical engineering and the design of safety components; and
law enforcement, sentencing, parole and predictive policing.
Where you use the service to produce material for the public in one of these fields, disclose that AI was involved. General wellness — sleep, exercise, nutrition, stress, habit-building — is not a high-risk field for the purposes of this section, and neither is your own private use in thinking a problem through.
Academic work has its own rule: follow your institution’s policy, and do not present AI-assisted work as unaided where that breaks it.
10The service itself
Do not:
attempt to defeat a safety measure, a content filter, a rate limit or a spending control, including by prompt injection or jailbreaking;
run adversarial or red-team testing against Switchboard or the models behind it without our written permission — ask us first at abuse@switchboardai.ca and we will usually say yes to serious researchers;
open multiple accounts, use false details, or share credentials in order to exceed an allowance or evade a suspension;
scrape, crawl or automatically extract data or output from the service, or resell access to it;
reverse engineer the service or attempt to derive its routing logic, except where the law expressly allows it;
place a load on the service that degrades it for other customers; or
use a VPN, proxy or false location to get around a geographic restriction imposed by us or by a provider.
11Models and provider rules
The models behind Switchboard belong to other companies, and their rules apply to you as well as to us. In particular, do not:
use output to train, fine-tune, distil or evaluate a competing AI model;
systematically extract model behaviour, weights or training data;
resell model access, or build a service whose substance is reselling it; or
breach the published terms of any provider listed on the providers page.
12How we enforce this
We do not read conversations for enforcement. What we have is automated screening of uploads and requests, the abuse reports people send us, signals from the providers we route to, and the audit log — which records that a call happened, not what was said.
When something surfaces, we respond in proportion: a warning, a narrowed feature, a suspension, or closing the account. Where the situation allows, we tell you what the problem is and give you a chance to fix it. Where it does not — child safety above all — we act first.
You can ask a person to look again. If an automated control has got you wrong, write to support@switchboardai.ca and a person will review it and reply. The exception is section 3, which we do not review.
If we close your account for breaching this policy, we do not refund the current period. We may report conduct to law enforcement or to a regulator where the law requires it or where someone is at risk.
13Reporting a problem
If you see Switchboard being used in a way that breaks this policy, tell us at abuse@switchboardai.ca. Include enough detail for us to find it. We take reports about children and about credible threats to someone’s safety first and fastest.
If you find a security vulnerability in Switchboard, please report it to the same address and give us a reasonable chance to fix it before telling anyone else. We will not pursue you for good-faith security research that respects other people’s data.
—How to reach us
Protocol 42 Inc. is a corporation incorporated in Ontario, Canada, and operates Switchboard at https://www.switchboardai.ca.