Skip to content
Takeida
Start free
LegalAcceptable Use Policy

Acceptable Use Policy

The shortest of the three, and the one most often cited at somebody. Every rule is numbered so it can be quoted in one line of an email, and every decision made under it can be appealed.

Effective
13 September 2026
Version
0.1
Status

Draft, pending legal review

Where it applies

Everything on the service. Hardest on Concepts you have made public.

What happens

A proportionate response, starting with a note. You are told which rule and why.

How to appeal

Within 30 days, to a different person from the one who decided.

The short version

The rules, and what happens when one is broken. Each rule below is numbered so it can be cited precisely.

  1. Do not publish what is not yours.

    Importing somebody else's work because it was readable is the most common way this rule gets broken, and being able to read something has never been permission to take it.

  2. Do not publish what harms people.

    Unlawful material, content that sexualises children, incitement, targeted harassment, malware. Section 2 numbers each one.

  3. Your Agent is you.

    Rate limits apply to automated access, and the member whose Agent it is answers for what it does.

  4. We respond proportionately.

    A note first, in most cases. Restriction, unpublishing, suspension and termination are a ladder, not a menu, with two exceptions that skip it.

  5. You are told which rule and why.

    Not a message saying a policy was violated. The rule, the material, and the decision.

  6. You can appeal, to someone else.

    Within 30 days, reviewed by a different person. If we were wrong, what we restricted is restored.

Contents10 sections

Where this applies

This policy applies to everything you do on the service, and it bites hardest on public Concepts, because those are readable by anyone with the address and we are the ones hosting them.

It is part of the Terms of Service. Breaking it is breaking those. It covers what you write, what you import, what you publish, and what your Agent does on your behalf.

Private Concepts are yours, and we do not read them looking for breaches. The rules still apply to them, and we act when something is reported to us or when we are required to, but there is no system scanning your organization's writing.

What may not be published

Eight rules. Each is numbered so it can be quoted precisely, which is the difference between a decision you can argue with and one you cannot.

  1. 2.1

    Do not publish material that is unlawful where you are or where we are.

    Including material whose publication is itself the offence. If it would be a crime to hand somebody a printed copy, it is a breach of this policy to put it on the trunk of a public Concept.

  2. 2.2

    Do not publish material that sexualises a child, in any form, including drawn or generated.

    This is the one rule with no proportionate response and no appeal to a lesser outcome. It is removed, the account ends, and it is reported to the authorities.

  3. 2.3

    Do not publish content that incites violence, or that threatens or targets a person.

    Writing about violence is not inciting it, and a document analysing an attack is not a threat. What this forbids is material meant to bring harm to somebody, and material that makes a specific person a target.

  4. 2.4

    Do not use a Concept to harass somebody.

    A public Concept is a durable, addressable, indexed record, which makes it an unusually effective instrument for this. Publishing a person's private details, or building a body of knowledge whose subject is degrading one individual, is a breach whether or not anything in it is false.

  5. 2.5

    Do not publish other people's personal data without a lawful basis.

    Customer lists, employee records, and documents holding somebody's contact details or health or financial information. Private Concepts are the right place for these; publishing one is a decision with consequences for people who never agreed to it.

  6. 2.6

    Do not publish malware, exploit payloads meant to be run against somebody, or stolen credentials.

    Security research, write-ups and proof of concept code published responsibly are not this. Credential dumps and material whose only use is to attack a system somebody else runs are.

  7. 2.7

    Do not impersonate a person or an organization.

    Presenting a Concept as the official record of an organization you do not represent, or writing under somebody else's name. Commentary, criticism and clearly labelled parody are not impersonation.

  8. 2.8

    Do not manipulate ratings, follows or adaptation counts.

    Multiple accounts rating the same Concept, arrangements to rate each other's work, or automated follows. Those numbers are the only public signal a reader has about whether something is worth their time, and a rating is one person's honest judgement or it is nothing.

Content that is not yours to publish

The most common breach of this policy is not malice. It is somebody importing a large amount of text that was never theirs to bring, usually because they could read it.

Import is the fastest way to move a very large body of writing into an organization at once, which makes it the most likely route for material somebody else owns. Do not import work that is not yours unless you hold permission from whoever owns it.

Where a Concept has been made adaptable, its owner has named the terms on which it may be taken, and those terms are what governs. Read them before you build on something, and honour what they require, which for three of the four options means crediting the original writers.

If you are unsure whether you may bring something in, the answer is to ask its owner, not to import it and find out. An import lands in a Draft and never on the trunk, so nothing enters the shared record without a person approving it. That is a safeguard for your organization, and it is not a defence against a copyright claim.

Agents, rate limits and scraping

An Agent connects through its own key. The member whose Agent it is answers for what it does, and the audit record names that person rather than the instrument.

  • Requests are rate limited. Do not work around a limit with more keys, more accounts, or more addresses. If a limit is genuinely in the way of something legitimate, write to us.
  • Do not scrape public Concepts in bulk. Reading one is what publishing invited; systematically taking all of them is the thing the adaptable setting exists to decide, and going around it is going around the owner.
  • An Agent belonging to somebody outside an organization is capped at writer standing. Do not attempt to obtain more than that, and do not lend a person's sign-in to a tool to sidestep the cap.
  • Do not point an Agent at a Concept whose owner has set the Agents policy to exclude it. That setting is checked on every request, so narrowing it stops work already under way, and treating that as an obstacle to route around is a breach of this policy.

Using an Agent to produce writing is not a breach of anything. The product is built for it, everything an Agent contributes is visibly attributed, and a Proposal from one carries no more authority than a Proposal from anybody else.

The service itself

Rules about the service rather than about content.

  • Do not attempt to reach data you were not given access to, or to work around a permission check. Access is checked on every request, and probing it is not research unless it is done under the terms in our security disclosure page, which is being published before general availability.
  • Do not attempt to degrade the service for other people, and do not resell capacity, access or storage you were given for your own use.
  • Do not use the service to send unsolicited email, or to host the landing content for something that does.
  • Do not create accounts to evade a suspension. An account made to get around a decision is closed on sight, and it converts a disagreement we might have resolved into one we will not.

Reporting something

Write to abuse@takeida.com with the address of the Concept, the document if you can name it, and which rule you think it breaks. Quoting a number from section 2 is the fastest way to be understood.

  1. Somebody reports it

    Any time

    By email, with the address of the Concept and enough detail for us to find the thing being complained about. Anonymous reports are read; reports we cannot locate cannot be acted on.

  2. We look

    Within 2 working days

    A person, not a classifier. We acknowledge the report and say whether we are investigating. Where the material is plainly unlawful we restrict access while we look rather than after.

  3. We decide, and say why

    Usually within 7 days

    The owner is told what was reported, what we decided and which rule it turned on. The reporter is told the outcome. We do not issue a decision that only says a policy was violated.

  4. The owner may appeal

    Within 30 days of the decision

    To a different person from the one who decided. An appeal that shows we got the facts wrong reverses the decision and restores what was restricted.

Reports made in bad faith, and reports used as a tactic against somebody, are themselves a breach of this policy. We notice patterns in who is reporting what.

What we do, in order

The response fits what happened. Most reports are a misunderstanding, or something somebody will fix once they know about it, and treating those the same way as a deliberate breach would be both unfair and useless.

  1. A note

    We tell you what we found and ask you to deal with it. Most reports end here.

  2. We restrict the thing

    One document or one Concept stops being reachable while it is sorted out. Everything else carries on.

  3. We unpublish it

    A public Concept becomes private. Your organization keeps it and can still export it.

  4. We suspend the account

    Sign-in stops. Export keeps working, because withholding somebody's own writing is a punishment we are not willing to impose.

  5. The account ends

    For repeated or deliberate breaches. The 30 day export window in the Terms still applies unless the content is unlawful to retain.

Two things skip the ladder. Material that sexualises children, and a credible, specific threat to a person. Both go straight to removal and termination, and both are reported to the authorities without waiting to hear from you.

Whatever we do, you are told what it was, which rule it turned on, and what the material was. The only exception is where telling you would itself be unlawful, which in practice means a law enforcement request that forbids it.

We do not delete your writing as a sanction. Restriction and unpublishing change who can reach something; they do not destroy it, and export keeps working throughout. The exception is material we are not lawfully permitted to retain.

Appealing a decision

Every decision made under this policy can be appealed, within 30 days, by replying to the message that told you about it or writing to appeals@takeida.com.

  • It is reviewed by a different person from the one who decided. An appeal to the same person is not an appeal.
  • We answer within 14 days, with the reasoning rather than the outcome alone.
  • If we got it wrong, what was restricted is restored, and the decision is removed from your record rather than left on it as a warning that was never warranted.
  • If we got it right, you are told which part of your appeal did not change our view, so a second attempt can address it rather than repeat itself.

An appeal does not stop a restriction while it is being considered, except where the material was restricted as plainly unlawful and it turns out on the face of the appeal that it was not.

Contact

Reports go to abuse@takeida.com. Appeals go to appeals@takeida.com. Copyright complaints go to legal@takeida.com. Security reports go to security@takeida.com.

A person reads each of these. Anything that does not fit is on the contact page.