Octuo AI Disclosure

This document explains exactly which AI providers Octuo uses, what data flows to each, and what the limits of AI output are. It is part of the legal documents set referenced in the Privacy Policy and Terms of Service.

1. Providers we use

Octuo uses large language models from the following providers. Your prompts (and the system context Octuo composes around them) are forwarded to whichever provider is best for your task or the one you selected.

We forward your prompt and only the system context Octuo composes around it (your model preferences, the chat history relevant to the current turn, the tool definitions available, and your selected agent capabilities). We do not forward:

• Your password.
• Vault plaintext (unless you explicitly dispense a credential into the prompt yourself).
• Other users' Content.
• Telemetry or analytics.

2. Choosing the model

By default, Octuo's smart router selects the model best matched to your prompt's apparent intent (cost-aware; capability-aware). You can override the choice in:

• The chat compose surface (per-message model picker).
• Account → Settings → Models (default model preference).
• Per-conversation pinned model.

3. Limits of AI output

AI output can be wrong. We say this in many places because it is the single most important thing to remember.

• Hallucination: large language models can generate plausible-sounding but factually incorrect content. They may fabricate citations, quotes, statistics, or legal / medical / safety information.
• Bias: models reflect the biases of the data they were trained on. Output may be skewed in ways that are not always visible.
• Context limits: models have a maximum context window. Long conversations or large attachments may be truncated.
• Date cutoff: each model has a knowledge cutoff date; events after that date may be unknown to the model unless we provide them in the context.
• No real-time access: by default the model does not browse the web, run code, or call external APIs. You enable specific tools per session.

Do not use Octuo for medical, legal, financial, or safety-critical decisions without independent professional verification.

4. Agent autonomy and the safety boundary

Beyond chat, Octuo offers agent capabilities that let the AI take actions on your behalf. The boundary you control:

• No agent capability is enabled by default for sensitive surfaces.
• Each capability you grant (file access, vault access, multi-device control) is a deliberate choice in Account → Settings or in the in-flow consent dialog.
• The audit log records every agent action for review.
• You can revoke a capability at any time; the revocation takes effect immediately for new actions but does not "undo" past authorized actions (the action has already been performed).

We provide best-effort sandboxing of agent actions but the underlying risk remains: you are authorizing the AI to act, and AI judgment is imperfect.

5. Data used to improve Octuo (NOT to train provider models)

Tutuo collects telemetry to improve Octuo:

• What features people use (anonymous aggregate).
• What errors occur (with redacted user identifiers when necessary).
• How Octuo's smart router decisions perform (whether the model picked produced a result the user accepted or revised).

This telemetry stays inside Tutuo. We do not send it to OpenAI, Anthropic, Google, or any other AI provider for their model training. See Privacy Policy §2.2 for the full list.

If you want to opt out of telemetry collection beyond what's strictly required to deliver the Service, contact privacy@tutuo.ai. (Some telemetry — e.g. audit log of authentication events — is required for security and cannot be opted out without deleting the account.)

6. Content moderation

We rely on the content-moderation infrastructure of the AI providers we route to. They may decline to respond to certain prompts or attach safety disclaimers. We do not currently add Tutuo-side content moderation on top of provider moderation.

7. Provider changes

We may add, remove, or change providers as the AI landscape evolves. If we add a new provider, this document will be updated and listed in the version history. If a provider materially changes their training-on-input policy in a way that affects your data, we will notify by email at least 30 days before changing routing defaults.

8. Third-party terms

By using Octuo's AI features, you accept the third-party providers' usage policies for prompts forwarded to them:

• OpenAI Usage Policies: https://openai.com/policies/usage-policies
• Anthropic Usage Policy: https://anthropic.com/legal/aup
• Google Generative AI Prohibited Use Policy: https://policies.google.com/terms/generative-ai/use-policy

We make reasonable effort to surface these policies in the in-app help, but you are responsible for ensuring your use complies.

9. EU AI Act and similar regimes

Octuo is a US-based service. We monitor EU AI Act developments and will adapt as relevant requirements come into force. As of this document's effective date, Octuo's classification under the EU AI Act is, to our reasonable assessment, "limited risk" (transparency obligations apply); we comply with the key transparency obligations by:

• Disclosing that you are interacting with AI (this document, the Beta Notice, the in-app branding).
• Marking AI-generated outputs as such (the chat surface labels each AI message).
• Not deploying any prohibited-risk applications (no biometric categorization, no social scoring, no manipulative dark patterns).

If you have a specific EU AI Act question, email legal@tutuo.ai.

10. Contact