Troi — AI Transparency Disclosure

Draft pending attorney review. This is a usable draft implementing Troi's AI-transparency decisions. It is not a substitute for licensed-counsel sign-off before publication.

Effective date: 9 July 2026 · Version: Draft 1.0 · Last updated: 9 July 2026

In short

• You're talking to AI. Many Troi features are powered by artificial intelligence. Responses may be inaccurate, incomplete, or out of date.
• AI is information, not professional advice. Nothing an AI feature produces is medical, mental-health, financial, tax, legal, or professional advice. Review it before you rely on it.
• You usually bring your own AI-provider key — your prompts go to your own account with that provider (often on your own Troi Runner), under your own settings. Any Troi-provided ("managed") AI uses a zero-retention / no-training tier.
• AI-generated images, video, and audio carry provenance marking (C2PA Content Credentials / audio watermark) so they can be identified as AI-made.
• You are in the loop. Coding agents propose changes for your approval by default, and AI-generated social posts are labeled as AI-generated to the platforms you publish to.
• Found an AI error? Tell us at info@kanah.app so we can review and, where appropriate, correct it.

This disclosure explains how Troi (operated by Kanah LLC) uses artificial intelligence, what that means for you, and your rights. It is written to meet the transparency duties of the EU AI Act (Article 50), California SB-942 and SB-243, and general consumer-protection law, and it applies globally. It sits alongside our Privacy Policy, Terms of Service, and Sub-processor List.

1. You are interacting with AI

Where you chat with an assistant or agent inside Troi, you are talking to a machine, not a human. We show this clearly:

• A persistent "AI" badge on chat and agent surfaces, plus a first-message banner: "You're talking to AI. Responses may be inaccurate."
• On voice surfaces, a spoken notice on the first utterance: "This is an AI voice…"

We do not try to pass AI off as a person, and we do not use AI to impersonate a specific real person without disclosure.

2. Which features use AI, and which providers power them

Troi uses third-party AI and media models to power its assistant, agent, generation, and media features. The current providers are:

Feature areas that use AI include, for example: the in-app assistants and agents; Troi Studio app-building, design, and coding-agent features; text, image, audio, and video generation; summarization and drafting; transcription and captions; and drafting content for publishing to external social platforms.

The current, itemized list of AI providers — with data categories and regions — is maintained on our Sub-processor List, which we update as providers change. We commit to 30 days' advance notice before adding a new Troi-operated sub-processor, including AI providers.

3. AI output can be wrong — and is not professional advice

AI models generate plausible-sounding text, code, and media that can be inaccurate, incomplete, biased, or entirely fabricated ("hallucinated"). Treat every AI output as a starting point you must review, not a finished or authoritative answer.

AI output is informational assistance only. It is not professional advice. In particular:

• Health & wellness: Troi Life is not a medical device and does not provide medical advice. Consult a qualified healthcare professional before acting on anything you record or read here.
• Finance: Troi tracks your own entries. We are not a registered investment adviser. This is not investment, financial, tax, or legal advice.
• Journal / mental health: Not therapy and not a substitute for professional mental-health care. If you are in crisis, contact your local emergency services or a crisis line (in the US, call or text 988).

These disclaimers also appear in-context on the relevant surfaces and on every AI reply in those workspaces. Troi does not make automated decisions about your employment, credit, or other high-stakes matters, and AI features are scoped to non-consequential assistance (such as drafting, templating, and summarizing) with your review.

4. Provenance marking on AI-generated media

When Troi generates media with AI, we mark it so it can be recognized as AI-made:

• Images and video: an embedded C2PA Content Credentials signed manifest, and the IPTC DigitalSourceType set to trainedAlgorithmicMedia.
• Audio (text-to-speech and any generated voice): an inaudible audio watermark.
• Provider watermarks preserved: where a provider embeds its own signal (for example SynthID), we do not strip it during processing.
• In-app label: AI-generated assets show an "AI-generated" label on playback.

Please do not remove, alter, or obscure these provenance markings.

5. How your data reaches AI providers — bring-your-own-key vs. managed

By default, you use your own AI-provider account. You configure a provider (such as Anthropic or OpenAI) and supply your own API key; your prompts are sent to your own account with that provider — often executed on your own Troi Runner — under your agreement with that provider. In this mode:

• That provider's retention and training settings — which you control — govern. Troi acts as a conduit and does not set them. If a no-training / zero-retention posture matters to you, select it in your own provider account.
• The AI provider is your processor under your terms, not a Troi sub-processor.

Temporary launch gateway. [[FOUNDER: remove this paragraph once the launch gateway is disabled — active < 1 week after launch]] For a brief period after launch, Troi may offer a free AI gateway (funded by Troi's own Microsoft Azure resources, routing to the OpenAI/Azure OpenAI API) so you can try AI features without your own key. During that period, requests are processed on a no-training basis (Azure/OpenAI do not train on API data by default) and pass through Microsoft/Azure, already listed in the Sub-processor List. This gateway is switched off shortly after launch, after which AI is bring-your-own-key only.

See the Privacy Policy for the full picture.

6. Human-in-the-loop for coding agents

Troi's coding agents can write and change code, but they are designed to keep you in control:

• By default, a coding agent proposes a change (a diff) and waits for your approval before it commits or pushes.
• Scheduled or automated agent runs do not silently push code on your behalf under the default setting.
• AI-authored portions of code may not be protected by copyright in some jurisdictions (including the US), and AI can surface code that is license-incompatible or infringing. If you enable more autonomous behavior, you do so at your own risk and remain responsible for reviewing what the agent produces.

7. Voice features

Troi's voice features are text-to-speech using preset library voices (via ElevenLabs), and carry the provenance and disclosure measures above. Troi does not clone voices and does not create a biometric "voiceprint" of you. You may upload a "voice profile" audio recording, which we store as your own content for use in features you enable — it is not turned into a biometric identifier. If Troi ever offers genuine voice cloning in the future, we would add an explicit consent flow and update this disclosure first. See the Privacy Policy.

8. AI labels on content you publish to social platforms

If you use Troi to publish content to external platforms (for example Instagram, TikTok, YouTube, LinkedIn, or X), and that content was created or materially edited with AI, Troi labels it as AI-generated to that platform using the platform's own AI-content flag (for example TikTok's AI-generated content flag, YouTube's altered/synthetic-content setting, Meta's AI-generated flag, and LinkedIn's AI flag). You control the "made for kids" and audience settings on your posts. Troi will not publish AI content about identifiable real people or political topics without a human-review attestation.

9. Your responsibilities

Because AI output can be wrong, you are responsible for reviewing and verifying anything you rely on or share — especially code you deploy, content you publish, decisions you make about your health or finances, and anything you send to other people. Do not use Troi's AI features to break the law, infringe others' rights, or generate prohibited content (see the Acceptable Use Policy / Terms of Service).

10. Report an AI error or request a correction

If an AI feature produced something inaccurate, harmful, or that misuses your data, tell us:

• Email: info@kanah.app (subject line "AI error report")
• Phone: +1 (424) 291-5470
• Or use the in-app feedback control on the relevant AI surface, where available.

Tell us the feature, roughly when it happened, and what went wrong. We review reports, use them to improve the Service, and — where the output involves your personal data — you also have the right to request correction (rectification) or deletion under the Privacy Policy and applicable law. We respond to such requests within 30 days.

11. Changes to this disclosure

We update this disclosure as our AI features and providers change. Material changes are announced in-app and/or by email, and every version is dated and archived.

Questions about this disclosure: info@kanah.app · Kanah LLC, 1603 Capitol Avenue, Suite 413A, #2280, Cheyenne, WY 82001, USA.