Kranth · Acceptable Use Policy

The full Terms of Service apply. This document is the specific list of things you can't do with Kranth.

Violating this policy can get your account suspended or terminated immediately, with no refund.

#You may not use Kranth to

#Generate or amplify illegal content

No CSAM. No sexualized content involving minors. No exceptions.
No content that incites violence against identifiable people or groups.
No content that violates IP law (counterfeit goods, pirated media, etc.) where Kranth is being used to draft or test the listing.
No content that violates laws applicable in your jurisdiction (controlled substances, weapons, etc.). Knowing those is your responsibility.

Personas are synthetic, with synthetic biases. They're not actually anyone.
You may NOT submit ideas designed to elicit reactions in the voice of a specific named real person ("how would Sam Altman react to my pricing page") without that person's consent OR a clear public-figure parody context (think: Saturday Night Live test).
Public figures get a narrow exception for legitimate journalism, parody, and political commentary, judged by the same standard a court would apply to a published article.

#Target harassment

No using Kranth to stress-test harassment campaigns against named individuals or small groups.
No using sim outputs to compose abuse, threats, or doxxing posts.
No using sim outputs to refine social engineering against specific people.
No publishing a sim or debate at a public share link as a vehicle for any of the above. Public links are part of the Service and this policy applies to them.
No presenting voice-debate audio (synthetic voices) as a recording of a real person.

#Interrogate someone who didn't agree to it, or grade them on it

Interview Mode points the panel at a person. That makes a set of misuses possible that the other modes don't have.

Don't put someone else in the chair without telling them. If a person is answering the panel, they must know they're talking to synthetic personas and that their answers are transcribed and stored. Running a session on someone through a shared screen or a hot mic without saying so is not allowed.
Don't use a readiness score to decide anything about a person. It is not a hiring, promotion, firing, admissions, lending, insurance, immigration, or tenancy assessment, and it must not be used as an input to one: not as a screen, not as a tiebreaker, not as "one signal among many". It measures how one person defended one decision on one occasion. Several jurisdictions regulate automated assessment of people; we're not going to be the tool that puts you on the wrong side of that.
Don't submit a third party's personal data as interview context unless you have a lawful basis to and would be comfortable telling them.
Don't present a transcript or verdict as an evaluation by a real expert. The panel has job titles because that makes the questions sharper. Nobody on it is real, holds a licence, or has reviewed anything.

#Interfere with elections

No simulating voter reactions to manipulate them.
No A/B-testing disinformation or content that incites suppression of legal voting.
No automating astroturf comment generation about candidates / referenda / political organizations.
Legitimate political communications from campaigns, journalists, or advocacy organizations are fine. The line is truthful intent.

#Workflow for fraud

Don't use Kranth to draft phishing copy.
Don't use Kranth to test which scam variants get fewer skeptical reactions.
Don't use Kranth to optimize fake-review generation.

#Try to break Kranth

No probing for security vulnerabilities outside the disclosure program. (Want to test? Email [email protected] and we'll authorize you.)
No attempting to access another org's data, recover hashed API keys, replay webhooks, or bypass rate limits.
No reselling raw API access without a partnership agreement.

#Resell without permission

The API is for your org's use.
White-label resale (your customers calling Kranth through your service, branded as your service) requires an Enterprise contract with a rev-share clause.
If you're building a tool ON Kranth that aggregates Kranth's signal with other signals, that's fine. It's a normal SDK integration.

Stress-test malware payloads.
Probe what content Kranth will pass through to LLM providers in an attempt to jailbreak them upstream.
Run sims at adversarial scale designed to exhaust LLM provider quotas (DoS-by-proxy).

#How we enforce

Every submitted idea / debate topic passes through an automated pre-run safety classifier with a narrow policy (CSAM, weapons instructions, terrorism, credible threats of violence, self-harm instructions). If it trips, the run is rejected before any credits are charged. Beyond that narrow check, we don't read your ideas. We also react to:

Reports ([email protected]). We investigate within 5 business days.
Pattern detection in the audit log when a customer's usage signal looks anomalous (sudden surge in similar prompts, ratio of failed sims, etc.).
Subpoena or court order. We cooperate with lawful requests.

Publicly shared content: if you publish a sim or debate at a public share link and it's reported for violating this policy, we may unpublish or remove it (in addition to any account-level action).

For violations we judge as severe (CSAM, election interference, targeted harassment with identified victims): immediate suspension + report to authorities where required.

For ambiguous cases: we'll email the account owner with what we saw and ask for context. Quick honest response usually clears it up.

#Right to refuse

We may decline to onboard or continue serving accounts that:

Operate in lines of business where we'd be uncomfortable having the service tied to (we'll tell you what those are if you ask).
Have a track record of AUP violations on other platforms we're aware of.
We have credible reason to believe will use Kranth for the prohibited purposes above.

This isn't a denial-of-service in bad faith. We reserve the right to decline business.

#Appeals

If we suspend your account and you think we're wrong:

Email [email protected] within 30 days with your account email and what you think happened.
We reply within 5 business days with our reasoning and what evidence (if any) we relied on.
If you disagree, you can escalate to [email protected]. A different person reviews.
After that, the terms' dispute-resolution section applies.

#Where the AUP doesn't reach

We don't moderate the LLM provider's own behavior. If you submit content that an LLM refuses, that's between you and that LLM provider.
We don't moderate the Kranth GitHub App's posted comments. The repo owner is responsible for what gets posted into their repo.
We don't moderate what you do with Kranth outputs after we hand them to you. Use them ethically; that's on you.

#DMCA

Designated agent for DMCA notices: [email protected]
We respond to valid DMCA takedowns by removing the offending content and notifying the account owner.
We honor counter-notices that meet 17 U.S.C. § 512 requirements.

#Reporting violations

Abuse / AUP report: [email protected]
DMCA notice: [email protected]
Security disclosure: [email protected]
Privacy / data protection request: [email protected]

The shorter version: if you'd be embarrassed to explain what you're doing with Kranth to a journalist, don't do it.

Acceptable Use Policy