Sparky
Our safety promise

If we wouldn't trust it with our own kids, we don't ship it.

Four commitments that shape every decision we make, from the conversation model to the stitch on the cube.

Describes a product we are building. Specifications, safety features, and certifications listed on this page are design targets, not yet certified. No Sparky device has shipped or been independently audited yet. Final, verified information will be published before the first device is sold. Questions: hello@sparky.ai.
Last updated: May 17, 2026

1. Content that respects a child's age

Sparky speaks like a kind, curious adult, not a chatbot showing off. Responses are shaped by an age-aware system prompt and filtered through a moderation layer trained on content guidelines for children's media. We test against red-team prompts on a weekly basis and adjust as new edge cases emerge. If a child asks a question that needs an adult answer (health, scary news, conflict at home), Sparky says so and gently encourages them to talk to a grown-up.

What you will not hear from Sparky

  • Adult content of any kind.
  • Violent, graphic, or scary descriptions, even if the child asks.
  • Marketing for third-party brands, products, or services.
  • Opinions on politics, religion, or other adult-domain topics. Sparky points the conversation to a parent.
  • Medical, legal, or financial advice. We explain what those are and recommend asking an adult.

2. Full transparency for parents

You can see everything your child says to Sparky and how Sparky replied. The parent app shows complete transcripts, flags moments worth reviewing (a difficult emotion, a question Sparky deferred), and lets you pause the device or the entire account in one tap. Weekly "what your kid asked about" summaries are sent on request, without ever surfacing them automatically.

3. Privacy as the default, not an upsell

We don't profile children. We don't run behavioral ads. We don't sell or rent data. Voice is processed to answer in the moment and is not used to train third-party AI models. The technical details are on the privacy page and the legal commitments are on the COPPA page.

4. Physical safety as a hardware spec

Hardware design targets for the Sparky device. None of the certifications listed below have been issued yet — they are the standards we are designing the device to meet before first ship.

  • Knit fabric cover designed to be washable and to meet children's product safety standards for fiber content and dye fastness.
  • Internal electronics designed to sit behind a snap-fit casing that requires an adult tool to open. No small loose parts.
  • Battery designed to be enclosed and non-removable by the user, following UN 38.3 transport standards.
  • Voice volume planned to be capped below WHO recommended listening levels at the device's native speaker distance.
  • Certifications we are targeting before first ship: CE (EU), FCC (US), UKCA, CPSIA (children's product safety, US), and EN 71 (toy safety, EU). We will publish issued certificates here once granted.

How to reach us

See or hear something that worried you? Tell us immediately: hello@sparky.ai. We treat every report as urgent and respond as quickly as we can.