A few honest limits, in plain language.
Cuttlefish is built to be reliable, private, and under your control. This page is the candid part: what we can and can't promise, how to treat what the AI gives you, and where the responsibility sits. It's worth two minutes.
Provided by Moguls Inc. Last updated March 25, 2026. Questions? Email [email protected] or [email protected].
What this disclaimer covers.
Each section below tells you one thing plainly: what we don't guarantee, and how to use Cuttlefish well anyway.
1. The software is provided “as is”
Cuttlefish is provided on an “as is” and “as available” basis, without warranties of any kind to the maximum extent permitted by applicable law.
In plain terms: we work hard to make Cuttlefish reliable and secure, but we can't promise it will be completely free of bugs, errors, or security vulnerabilities. Use it with that understanding, and keep good habits for anything important.
2. AI output is not professional advice
This is the most important thing on the page. Content an AI model produces through Cuttlefish is not a substitute for professional advice of any kind. In particular, don't rely on AI output as:
- Legal advice — an AI model is not a licensed attorney and cannot give you legal counsel.
- Medical advice — an AI model is not a healthcare professional and cannot diagnose or treat a condition.
- Financial advice — an AI model is not a licensed financial advisor and cannot make investment recommendations for you.
- Engineering or safety specifications — AI output should not be used as final engineering, architectural, or safety-critical specifications without review by a qualified professional.
- Security guarantees — AI-generated code or security settings should always be reviewed by a qualified person before you rely on them in production.
AI models can produce output that sounds confident and authoritative while being factually wrong — sometimes called a “hallucination.” Always verify anything important independently before you act on it. Cuttlefish gives you a clear record of what the AI did so it's easier to check.
3. Guardrails and governance have limits
Cuttlefish includes guardrails: the controls that hold the AI to the permissions you've granted, ask before consequential actions, and keep a tamper-proof record of what happened. We think this is a meaningful layer of protection, and it's a real reason people trust Cuttlefish. It's also honest to be clear about what it can and can't do:
- It's a strong layer, not a guarantee. Guardrails are one part of keeping you safe. No control system can catch every possible misuse or every bad output.
- Coverage is written by people. The rules behind the guardrails are designed by humans and may not anticipate every edge case. New risks and failure modes are discovered over time.
- Risk is reduced, not removed. Guardrails lower the chance of something going wrong; they don't eliminate it. Use your own judgment and add safeguards that fit your situation.
- The landscape keeps changing. AI safety moves quickly. We update the guardrails regularly, but there can be a gap between a brand-new risk and an updated rule that addresses it.
4. Third-party AI providers
Cuttlefish lets you connect to outside AI providers — such as OpenAI, Anthropic, Google, and Mistral — using your own account and keys. Moguls Inc is not responsible for:
- The accuracy, quality, or safety of output generated by a third-party AI model.
- Interruptions, outages, or slow performance on a third-party provider's side.
- Changes a provider makes to its pricing, terms, or availability.
- How a provider handles your data once your prompt leaves your machine and reaches them.
- Content a provider flags or refuses under its own content rules.
Your use of any outside AI provider is governed entirely by your agreement with that provider. We strongly recommend reading the terms of service and privacy policy of any provider you connect through Cuttlefish. If you'd rather keep an exchange fully on your machine, Cuttlefish also runs private, local models so prompts and data never leave your computer.
5. Local data and data loss
Cuttlefish is local-first. Most of your data — conversations, files, settings, and configurations — lives on your own machine. That's good for privacy, and it also means a few things are on you:
- Backups are your responsibility. We don't automatically copy your local data to the cloud. If your drive fails, local data can be lost.
- Your computer's health matters. Crashes, power failures, disk corruption, or operating-system updates can cause data loss. Moguls Inc isn't responsible for loss caused by your local environment.
- Uninstalling. Removing Cuttlefish may also remove its local data, depending on your operating system and how you uninstall.
Keep regular backups of anything important on your machine, independent of Cuttlefish, the same way you would for any other application.
6. Experimental and beta features
From time to time Cuttlefish includes features marked “experimental,” “beta,” or “preview.” When you see one of those labels, treat the feature differently from the rest of the product:
- It may be unstable, incomplete, or change without notice.
- It may be removed entirely in a future release.
- It carries additional risk beyond the limits described elsewhere on this page.
Use experimental features at your own discretion, and don't rely on them for work you can't afford to redo.
7. Questions about this disclaimer
If anything here is unclear, get in touch — we'd rather you ask.
Moguls Inc
Legal: [email protected]
Support: [email protected]
Admin: [email protected]
Web: getcuttlefish.app
You can also read the privacy policy, the terms of use, and how we approach security, or download Cuttlefish to try it for yourself.