AI Disclosure & Transparency

Version 1.6 · Effective July 19, 2026 · Last updated July 20, 2026

1. You are interacting with AI

Some features of this Service use artificial intelligence — including generative AI — to produce responses, content, or recommendations. Where such a feature responds to you directly (e.g. a chat interface), you are interacting with an automated system, not a human, unless the interface explicitly says a human is involved. This disclosure is provided consistent with AI chatbot-disclosure laws (e.g. California's Bot Disclosure Law, Cal. Bus. & Prof. Code §17941) and the EU AI Act's transparency obligations for systems that interact with natural persons (Article 50).

2. What AI features do here

Sound It Out uses a constrained GPT-5.6-family model to draft short phonics practice passages and a brief, encouraging coaching lead-in. The model receives only an exact, hand-curated word bank and selected phonics level—not a learner name, microphone audio, progress data, or free-form child prompt. A parent may choose one fixed practice-theme category; the server converts it to application-authored guidance and the child cannot enter prompt text. Application code then tokenizes the draft and rejects any word outside that bank or the small heart-word allowlist. Moderation, daily request caps, and a curated fallback provide additional protections. A failed draft is never shown to a learner. The app's sound-by-sound coaching sequence is deterministic application logic, not a free-form AI answer. Its versioned instructional audio was pre-generated from explicit phoneme and word prompts with the open-weight Kokoro-82M voice model; the model does not run in the browser or receive learner data.

3. Who processes your input

When you use an AI feature, the constrained prompt and generated response are processed by our third-party AI provider(s): OpenAI. OpenAI states that API inputs and outputs are not used to train or improve its models by default unless an API customer explicitly opts in. OpenAI may retain API abuse-monitoring logs for up to 30 days by default. See OpenAI's API data-controls documentation and our Privacy Policy for details about the limited data sent.

4. Limitations — please read

• AI output can be inaccurate, incomplete, outdated, or biased. It reflects patterns in training data, not verified facts.
• AI output is not professional advice — not legal, medical, financial, tax, or safety advice — even if it's phrased confidently or in a professional tone.
• You're responsible for independently verifying anything you rely on before acting on it.
• AI-generated content may unintentionally resemble existing copyrighted or trademarked material; we don't guarantee it's free to use for every purpose.

5. High-risk & consequential uses

This Service's default AI disclosure and safeguards are not sufficient if you use (or let your customers use) an AI feature to make or materially inform a decision about a specific person in an area such as employment, credit/lending, housing, insurance, healthcare, or education. Depending on your jurisdiction and use case, that can trigger additional obligations, for example:

• The Colorado AI Act (SB 24-205) — impact assessments, consumer notice, and an opportunity to correct/appeal for "high-risk" AI systems used in consequential decisions.
• NYC Local Law 144 — bias audits and candidate notice for automated employment-decision tools used by NYC employers.
• The EU AI Act (Regulation (EU) 2024/1689) — Annex III "high-risk" obligations (risk management, human oversight, logging) if the system is offered to users in the EU.
• Illinois, and a growing list of other states, regulate AI use in specific contexts like employment interviews and insurance underwriting.

If you plan to use AI for consequential decisions, do not rely on this baseline disclosure alone; implement feature-specific controls, human review, and any required assessments before deployment.

6. Synthetic content

The service generates text passages and uses pre-generated synthetic voice clips for letter sounds, words, and sound games. It does not generate images or video. Optional read-aloud recognition uses the browser or operating system's speech service. Depending on the browser, that provider may process microphone audio on its servers. Sound It Out does not receive or store the audio and does not send it to OpenAI or Kokoro. The recognized text is used transiently to compare the passage; Sound It Out stores only the limited word-level practice events described in the Privacy Policy, not a complete transcript.

7. Your choices

If no OpenAI API key is configured or generation is unavailable, signed-in reading practice uses a hand-verified curated passage. The public homepage walkthrough is a fixed example that does not call OpenAI or save learner progress.

8. Contact

Questions about our use of AI can be submitted through Support.