AI Policy

What This Policy Covers

This Responsible AI Policy is published by Ocranthor LLC, a Florida limited liability company ("Ocranthor", "we", "us", "our"), and applies to every product, service, brand, and internal tool that uses artificial intelligence ("AI") and is operated by Ocranthor or one of its brands, including Solyx, Breed Production Studio, Ember & Tide, Bella Tavola Ristorante, Valdencross Academy, The Mountainview Deer Steak Joint, Nakhit Loobnan, Maison Fumoir, Crossroads Kitchen, Yakimono Japanese Restaurant, and Pure Heart Publishers (the "Brands").

It complements, and does not replace, the Privacy Policy, Terms of Service, Cookie Policy, and any in-app End User License Agreement. Where this AI Policy imposes a stricter standard than another Ocranthor document, the stricter standard governs.

"AI" in this policy means any system that produces text, images, audio, video, code, recommendations, classifications, or decisions using machine-learning models — including large language models (LLMs), diffusion and image-generation models, speech models, and computer-vision models.

Our AI Principles

• Human first. AI assists people; it does not replace human judgment in decisions that materially affect a person's life, health, finances, education, employment, or legal status.
• Transparency. When you interact with AI, you will be told. When content is generated by AI, we will say so clearly.
• Data minimization. We send the smallest amount of information needed to produce a useful result, and we do not retain it longer than necessary.
• No training on your data. Your prompts, files, photos, voice, and outputs are never used to train, fine-tune, or improve any AI model — ours or a third party's.
• Safety by default. Safety filters, content moderation, and abuse safeguards stay on in production. We do not ship features that require turning them off.
• Accountability. A human at Ocranthor is responsible for every AI feature we ship. Mistakes are owned, disclosed, and fixed.
• Lawful and rights-respecting. We follow applicable AI, privacy, copyright, and consumer-protection laws in the jurisdictions we serve, including the EU AI Act, the GDPR / UK GDPR, U.S. state privacy laws, COPPA, and FERPA where relevant.

Where AI Is Used in Our Products

The table below summarises where AI currently appears in Ocranthor products. Any new AI feature will be added to this table at the time it ships.

Brand-specific AI uses (for example, future Valdencross Academy tutoring tools) will be governed by an additional brand-specific AI notice before the feature is made available to the public, and never without verifiable parental consent where children are involved.

How Your Data Flows Through AI Features

For full data-handling details, see the Privacy Policy. The high-level summary specific to AI is:

• In transit, not at rest. Prompts and the content you provide are transmitted to the model provider only for the time needed to produce a response. We do not store your prompts or model outputs on Ocranthor servers, except where you explicitly save them (for example, a conversation you choose to keep on-device, or media you choose to export).
• Model providers. AI inference is performed by vetted third-party providers operating under enterprise / API terms that prohibit training on customer data. As of the date above these providers include Google Cloud (Vertex AI — Gemini, Veo), OpenAI (GPT and image models), Anthropic (Claude), Black Forest Labs (Flux), Recraft, and FASHN AI. The current list is maintained in the Privacy Policy and updated when it changes.
• On-device when feasible. Where a task can be performed locally (for example, document scanning, financial categorization, on-device redaction of sensitive identifiers), we keep it on your device.
• Redaction before transmission. Where AI features touch sensitive categories (medical, financial), we redact direct identifiers on-device before any data leaves your device, to the extent technically feasible for the feature to function.
• No advertising use. We do not use AI — or the data you give to AI — to build advertising profiles, to perform cross-context behavioral advertising, or to sell or share personal information under the meaning of CCPA/CPRA, the Colorado Privacy Act, or similar U.S. state laws.

We Do Not Train Models on Your Data

This applies whether you are a free user, a paid subscriber, or an enterprise customer. We require contractual no-training and no-data-retention commitments from every model provider we route requests to, and we use the enterprise / zero-retention configurations of those providers where available.

We also do not sell, license, or otherwise make your data available to any party for the purpose of training AI models.

Human Oversight & Accountability

No AI system operated by Ocranthor makes a solely automated decision that produces a legal effect, or a similarly significant effect, concerning you (Art. 22 GDPR). In particular:

• We do not use AI to grant or deny credit, employment, housing, insurance, education enrollment, or government benefits.
• We do not use AI for biometric identification of natural persons in publicly accessible spaces, social scoring, predictive policing, emotion inference in workplaces or education, or any other practice prohibited under Article 5 of the EU AI Act.
• For any feature that could foreseeably influence a high-stakes decision about a person (for example, medical interpretation, legal drafting, financial coaching), the AI is advisory only, the user remains the decision-maker, and that limitation is stated in-product.

A named Ocranthor staff member is the responsible owner of each AI feature. Internal AI changes that materially affect users are reviewed against this policy before release.

Limitations & Disclaimers

• Not professional advice. AI features in our products do not provide medical, legal, financial, tax, psychological, or other regulated professional advice. Always consult a qualified professional for decisions in those areas.
• Hallucinations. Generative models can produce confident-sounding statements that are factually wrong or invented. Verify anything you intend to rely on.
• Currency. Models have a knowledge cutoff and may not reflect recent events or changes in law.
• Output ownership. Subject to the Terms of Service, you are responsible for how you use AI output, including for ensuring your use does not infringe third-party rights or violate applicable law.

Prohibited Uses of Ocranthor AI

You may not use Ocranthor AI features, and we will not knowingly enable any third party to use them, to:

• Generate sexual content involving minors, or any content that sexualises a real person without their explicit, contemporaneous, and revocable consent;
• Generate non-consensual intimate imagery (NCII) of any real person, including "deepfake" nudes;
• Generate content that impersonates a real person in a way intended to deceive (fraud, political disinformation, defamation, identity theft);
• Generate content depicting credible threats of violence, instructions for weapons of mass destruction, or material designed to facilitate terrorist acts;
• Produce malware, exploitation code, phishing kits, or instructions designed to defeat security controls of systems you do not own or are not authorised to test;
• Conduct surveillance, profiling, or behavioral scoring of individuals without a lawful basis and clear notice;
• Make automated decisions of legal or similarly significant effect about individuals (see Human Oversight);
• Circumvent age gates, content filters, watermarks, or provenance signals;
• Infringe third-party intellectual property, including by passing off generated content as the work of a specific living artist whose style was deliberately mimicked;
• Train a competing AI model on outputs obtained from our services.

We reserve the right to suspend or terminate access for any user who violates these prohibitions, and to cooperate with lawful requests from competent authorities in cases involving credible harm.

Children, Health, Finance, and Other Sensitive Contexts

Children. Our public products are not directed to children under 13. Where a future product (notably Valdencross Academy) is intentionally designed for minors, we will obtain verifiable parental consent before any AI processing of a child's data, will limit data flows to providers that contractually support educational and children's-data restrictions, and will publish a child-specific AI notice. We will not use AI to perform emotion inference or behavioral profiling of children.

Health. Solyx may help you organise health-related information you choose to enter, but it is not a medical device, does not diagnose, and does not replace a clinician. Health data you mark as such is handled with the additional safeguards described in the Privacy Policy — Medical section.

Finance. Solyx's financial coaching is informational only. Categorical totals are processed by AI to surface patterns; transaction-level data stays on your device. Nothing presented constitutes investment, tax, or accounting advice.

Audio & voice. Voice features process audio for the requested task (for example, transcription) and discard audio afterward. We do not perform voice-print identification or emotion analysis.

Biometrics. We do not use AI for biometric identification or categorization of natural persons.

Marking AI-Generated Content

To support transparency online, AI-generated images and videos produced by Ocranthor products carry, where technically supported by the underlying model provider, embedded provenance signals (for example, C2PA Content Credentials or provider-side watermarks). You should not rely on the absence of such a marker to conclude that content is human-made.

When AI is used to communicate with you in a way that could reasonably be mistaken for a human (for example, chat interfaces), the product makes the AI nature of the interaction clear, in line with Article 50 of the EU AI Act.

Bias, Fairness, and Quality

AI models reflect the data they were trained on and can produce biased, stereotyped, or culturally narrow output. We work to reduce avoidable harms by:

• Choosing model providers with documented safety and bias evaluations;
• Testing user-facing prompts against a range of names, languages, regions, and demographic descriptions;
• Publishing in-product guidance when a feature has known weaknesses (for example, low-resource languages);
• Providing a one-tap channel to report biased or harmful output (see Contact).

We do not claim our AI features are unbiased. We do commit to investigating credible reports and fixing problems we can fix.

Security of AI Pipelines

• All requests to model providers are made over TLS 1.2+ to provider-issued endpoints.
• Provider API keys are held in Firebase Secret Manager / Google Cloud Secret Manager and rotated on staff offboarding and on suspected compromise.
• Prompts and outputs are not logged by Ocranthor in a form that links them to your identity beyond what is strictly necessary for abuse prevention and operational debugging, and any such logs are retained for the shortest period operationally feasible.
• We monitor for prompt-injection, jailbreak, and exfiltration attempts against features that combine AI with tool use or document context, and we restrict the actions an AI feature can take on your behalf.

Your Rights & Controls

In addition to the rights described in the Privacy Policy, you specifically have the following AI-related rights:

• Right to know. You can request a description of the AI used in any Ocranthor feature you have used, including the categories of model and the categories of data sent.
• Right to opt out of optional AI features. Optional AI features (for example, AI-assisted suggestions, location context, AI coaching) can be disabled in-product without losing access to non-AI functionality.
• Right to human review. If an AI-influenced decision in our products affects you and you disagree with it, you can ask a human at Ocranthor to review it.
• Right to flag harmful output. You can report unsafe, biased, or unlawful output to ai@ocranthor.net; we triage these reports on a priority track.
• Right to deletion. Any AI-related data we hold about you that falls under the Privacy Policy can be deleted on request, subject to legal retention requirements.

AI Incident Response

We treat the following as AI incidents and act on them on a priority basis: a model producing content that materially harms a user, an unintended disclosure of user data through an AI feature, a successful prompt-injection or jailbreak that exposes restricted capability, a provider security event affecting our pipeline, or a regulator inquiry about an AI feature. Our process is:

• Contain — disable the affected feature or route as needed;
• Investigate — determine root cause and the population affected;
• Notify — inform affected users and, where required, regulators within applicable legal deadlines (for personal-data breaches under the GDPR, within 72 hours);
• Remediate — ship the fix, document it, and update this policy if the incident reveals a gap.

Changes to This Policy

We will update this policy when we add, materially change, or retire an AI capability, when a model provider changes in a way that affects you, or when applicable law changes. The "Last updated" date at the top of this page reflects the most recent revision. Where a change materially expands the categories of data processed, we will provide reasonable advance in-product notice before it takes effect.

Contact & Reports

For questions about this AI Policy, to request human review of an AI-influenced outcome, or to report harmful AI output:

• AI & safety: ai@ocranthor.net
• Privacy & data rights: privacy@ocranthor.net
• General: hello@ocranthor.net

Postal: Ocranthor LLC, Florida, USA — address on file with the Florida Division of Corporations and available on request.