Acceptable Use Policy

This Acceptable Use Policy ("AUP") defines the ground rules for using Dyva. It applies to dyva.ai, all Dyva apps, the Marketplace, the API, integrations (Discord, Telegram, Slack), embedded experiences, and every other service we operate (collectively, the "Service") — run by Dyva, Inc. ("Dyva," "we," "us").

Everyone who touches the Service is bound by this AUP. Creators, users, developers, API consumers, white-label partners — all of you. If you use Dyva, you follow these rules. If you disagree with any of them, stop using the Service.

This AUP is part of our Terms of Service. Where there is a conflict between this AUP and the Terms, the Terms win unless this AUP explicitly says otherwise.

Dyva is built on a simple premise: create interesting AI experiences without ruining it for everyone else. These baseline conduct expectations apply everywhere on the platform:

• Be honest about who you are. Do not impersonate Dyva staff, other users, or organizations. Do not misrepresent your identity, credentials, or affiliations to gain trust or access.
• Respect other people. No harassment, bullying, doxxing, stalking, or sustained unwanted contact. This applies to interactions with real people and when discussing real people through AI characters.
• Do not break things. No unauthorized access, hacking, credential stuffing, denial-of-service attacks, injection attacks, or exploiting vulnerabilities. Do not put unreasonable load on our infrastructure.
• Do not game the system. No fake accounts for extra free-tier allocations, no manipulating billing, no coordinated inauthentic behavior, no spam. If something feels like cheating, it probably is.
• Follow the law. Do not use Dyva for anything illegal in your jurisdiction or ours. We cooperate with law enforcement when required and when the situation warrants it.

These content rules apply to everything on Dyva: character configs, descriptions, conversations, Marketplace listings, profiles, feed posts, DMs, group messages, and AI-generated outputs.

• Child sexual abuse material (CSAM). Zero tolerance. Any content that sexually exploits or endangers minors — visual, written, or AI-generated — is immediately removed and reported to NCMEC and law enforcement. No exceptions. No warnings. Your account is terminated on the spot.
• Violence and threats. No content that threatens, incites, glorifies, or promotes violence against any person or group. This includes credible threats, instructions for carrying out violence, and graphic depictions of extreme violence intended to shock or recruit.
• Non-consensual intimate imagery. No intimate, sexual, or nude imagery of anyone shared without their consent. This explicitly includes AI-generated deepfakes and manipulated media depicting real people.
• Hate speech. No content promoting hatred, discrimination, or violence based on race, ethnicity, national origin, religion, gender, gender identity, sexual orientation, disability, age, or other protected characteristics.
• Harassment and bullying. No targeted campaigns to harass, intimidate, or demean individuals or groups. This covers coordinated abuse, doxxing, and weaponizing AI characters against real people.
• Self-harm and suicide. No content that promotes, glorifies, or provides instructions for self-harm or suicide. No depicting self-harm as desirable or discouraging professional help. If you are in crisis: 988 Suicide & Crisis Lifeline — call or text 988.
• Terrorism and extremism. No content supporting, promoting, or recruiting for terrorist organizations or violent extremist groups. No propaganda, recruitment material, operational planning, or glorification of terrorist acts.
• Illegal activity. No content that promotes or provides instructions for illegal activities — drug trafficking, weapons manufacturing, human trafficking, fraud, money laundering, or anything else that would land you in court.
• Unauthorized personal information. Do not share anyone's private information without their consent: home addresses, phone numbers, financial details, government IDs, medical records, login credentials, or anything else that belongs to them and not you.

Dyva is an AI platform, and AI introduces unique risks. These rules exist specifically because of that:

• No jailbreaking. Do not attempt to bypass, disable, or circumvent AI safety systems, content filters, or character boundaries. This includes prompt injection, system prompt extraction, adversarial inputs designed to make characters violate their constraints, and sharing techniques for doing so.
• No bypassing safety measures. Our safety systems exist for a reason. Do not attempt to manipulate AI characters into generating prohibited content through creative prompting, roleplay framing, hypothetical scenarios, or any other workaround. If a guardrail blocks something, that is the intended behavior.
• No impersonation of real people without consent. Do not create AI characters that impersonate real people — public figures or private individuals — without their verifiable consent. Parody and satire must be clearly labeled. We remove any character that could reasonably be confused for a real person's endorsement or participation.
• No fraudulent professional advice. Do not create characters that present AI output as genuine professional advice in medicine, law, finance, therapy, or any regulated field where bad advice causes real harm. Characters touching these topics must clearly state they are AI and not a substitute for licensed professionals.
• No disinformation. Do not use AI characters to create, amplify, or spread false information to deceive people, manipulate public opinion, interfere with elections, or cause harm. This includes deepfakes designed to mislead about content authenticity.
• No training competing models. Do not use Dyva outputs to train, fine-tune, or develop competing AI models without our written consent. Systematically collecting AI responses to build training datasets for third-party models is a violation.
• Say it is AI where people could think otherwise. If you put a Dyva character somewhere its audience could reasonably take it for a person — an embed, an integration, a bot in someone else's space — you are responsible for making the AI identity clear in that surface. The platform does not inject a disclosure into embedded or API-delivered responses, so it will not be said unless you say it. Passing a character off as a human being is a violation whether or not any specific message was a lie.

For developers using the Dyva REST API and WebSocket connections, these rules apply in addition to the API Terms of Service:

• Respect rate limits. Follow all rate limits and quotas documented in the API docs and response headers. Automated retries must use exponential backoff. Sustained attempts to exceed limits will get your API access revoked.
• Secure your keys. API keys are your responsibility. Never embed them in client-side code, public repos, or anywhere they could be exposed. If a key is compromised, notify us immediately at [email protected]. You own all usage and charges on your key, authorized or not.
• No unauthorized resale. Building products that use API outputs for your users is fine. White-labeling or reselling the API itself requires a separate agreement. Contact us if that is your use case.
• Handle data responsibly. All data from the Dyva API must be handled in compliance with applicable data protection laws (GDPR, CCPA, etc.). Do not use API data to build user profiles or track individuals beyond the scope of your authorized use.
• No billing circumvention. Do not exploit billing mechanisms, free-tier limits, or promotions. Multiple accounts for extra free-tier allocations, request manipulation to reduce billing, or exploiting billing errors all result in termination.

The Marketplace is where creators publish AI characters and sell what they build. The revenue split, payout rules, and eligibility bar are in the Creator Agreement; this section is about conduct. With that opportunity comes responsibility:

• Your work must be original. Do not publish characters that copy or substantially replicate another creator's work without permission. Every listing must be original or properly licensed. We process takedown requests from original creators promptly.
• Be honest in your listings. Descriptions must accurately reflect the character's behavior, capabilities, and limitations. No exaggerated claims to drive subscriptions. Previews and samples must represent the actual experience.
• No rating manipulation. Do not artificially inflate ratings, reviews, or download counts. This includes buying reviews, self-reviewing from alt accounts, coordinated rating campaigns, incentivizing positive reviews, and leaving fake negative reviews on competitors.
• No spam listings. Do not flood the Marketplace with duplicate, low-quality, or auto-generated listings. Every character should offer distinct value. We remove spammy listings without warning.
• Follow all content standards. Every Marketplace listing must comply with Section 3 above, the Terms of Service, and the Creator Agreement. Violations mean removal and possible enforcement action under Section 9.

Dyva generates speech and media — voice conversations, speech-to-text and text-to-speech, image generation, comics, and video generation, plus voice cloning where that feature is available to you. These rules apply to every one of them, and they apply whether or not a given feature is switched on for your account today:

• No deepfakes. Do not use Dyva's voice, image, or video generation to create deceptive media that impersonates real people without their consent. This includes cloning someone's voice, generating their likeness, or creating synthetic media designed to mislead others about who is speaking or appearing.
• No non-consensual voice cloning. You may only clone voices you have explicit permission to use. Cloning a public figure's voice, a friend's voice, or anyone else's voice without their documented consent is a violation. When you upload voice samples, you are certifying you have the right to use them.
• Generated media follows content standards. All AI-generated images, audio, video, and comics are subject to the same content standards in Section 3. The fact that something was AI-generated does not exempt it from the rules.
• No misleading synthetic media. Do not present AI-generated media as real footage, real audio, or real photography when the context could mislead. If it is synthetic, it should be identifiable as synthetic.

Dyva has social features — DMs, groups, feeds, and integrations with platforms like Discord, Telegram, and Slack. These spaces are subject to the following:

• DMs are not a loophole. All content standards apply in direct messages. Private does not mean unmoderated. Harassment, CSAM, and other prohibited content in DMs is treated exactly the same as public violations.
• Groups must be managed. Group creators are responsible for the content in their groups. If a group becomes a hub for prohibited content and the creator takes no action, the group gets removed and the creator faces enforcement.
• Feed content is public. Posts on the feed are visible to others and must meet all content standards. No spam, no harassment, no prohibited content. We remove violating posts and take action against repeat offenders.
• Integrations carry the same rules. Using Dyva through Discord, Telegram, Slack, or any other integration does not change the rules. Our AUP applies regardless of where or how you access the Service.
• No coordinated abuse. Do not organize groups, channels, or campaigns to harass individuals, manipulate platform features, or coordinate violations of this AUP. Coordinated inauthentic behavior results in action against all participating accounts.

Violations reach us three ways: automated classification on content as it is created and published, reports from users, and our own review. Automated detection is not comprehensive and we do not read everything — a violation that has gone unnoticed is not a violation we have permitted, and time elapsed is not consent. When we find one, here is what action looks like:

• Warning. A formal notice identifying the violation and requiring corrective action. Warnings are documented and weigh into future decisions.
• Content removal. Violating content is removed or access is disabled, with or without prior notice. Where feasible, we tell you what was removed and why.
• Temporary suspension. Account access suspended for a defined period. During suspension, you cannot access your account, publish content, or use the API. Subscription billing may continue unless stated otherwise.
• Permanent termination. Account permanently closed and all access revoked. For creators, this means removal of all Marketplace listings and forfeiture of unpaid revenue below the minimum payout threshold. Reserved for severe or repeated violations.
• Law enforcement referral. Illegal activity gets reported to law enforcement. We cooperate with investigations and comply with subpoenas, court orders, and search warrants.

The action we take depends on severity, frequency, and history. We do not need to warn you first, especially when safety or legal obligations are at stake. Appeals are available — see our Trust & Safety page for the process.

If you see something that violates this AUP, report it:

• In the product. There is a report control on user profiles, character profiles, individual chat messages, feed posts, scenes, and video pages. A report records what you reported, the reason you chose, and anything you wrote, and it enters the Trust & Safety queue as pending. You can see the reports you have filed, and their status, on your Safety page.
• Email. Send detailed reports to [email protected].

When reporting, include as much as you can:

• What the violation is and which rule it breaks
• The username or profile URL of the account involved
• Links to or screenshots of the violating content
• When it happened (date and approximate time)
• Any additional context that helps us evaluate it

Reports are confidential. We do not reveal who reported to the reported party, except where law requires it. We do not tolerate retaliation against good-faith reporters.

We may update this AUP at any time. Material changes get a date update and notification through the Service or email. Continued use after changes means acceptance. For questions, reach us at [email protected].