Responsible AI at EmberKeep

AI should support people, not replace them.

Artificial intelligence can be useful. It can also be wrong, biased, resource-intensive, overused, and deployed in places where human judgment belongs.

We don't think responsible use means pretending those concerns don't exist.

At EmberKeep, our approach is to use AI deliberately: for specific tasks where it can meaningfully help people, with clear boundaries around what it should not do.

For Beacon, that starts with a simple principle:

The goal isn't to automate teaching. It's to automate the work that keeps teachers from teaching.

Educators remain the decision-makers

Beacon uses AI to assist educators with certain tasks, such as organizing information, generating suggested materials, interpreting educator-provided information, and reducing repetitive administrative work.

AI does not know a student the way their educators and families do.

Beacon is therefore designed as a support tool—not an autonomous decision-maker.

AI-generated content should be reviewed by an educator. Beacon should not independently determine a student's eligibility, placement, services, discipline, diagnosis, grades, or other high-impact educational outcomes.

AI can assist. Educators decide.

AI should earn its place

We don't believe every feature becomes better simply because AI is added to it.

Many things software does well—calculations, data storage, progress tracking, charts, scheduling, access controls, and other deterministic tasks—do not inherently require generative AI.

Our goal is to use AI where its language or reasoning capabilities provide meaningful value rather than making every interaction “AI-powered.”

That distinction matters for accuracy, privacy, cost, efficiency, and environmental impact.

Student information deserves restraint

Educational technology can involve sensitive information about real children. That deserves particular care.

We aim to minimize unnecessary student information and use information for the features educators request.

Beacon is not designed to replace educators' responsibilities for determining what student information they are authorized to enter, upload, or share.

We do not sell student educational information.

AI can be wrong

Generative AI can produce inaccurate, incomplete, biased, or inappropriate information—even when it sounds confident.

We don't want Beacon to disguise that limitation.

AI-generated suggestions are starting points or assistance, not independently verified facts. Educators remain responsible for reviewing generated content before relying on it professionally.

Environmental responsibility

AI has a physical footprint. We don't pretend otherwise.

The computing infrastructure behind artificial intelligence uses energy, water, and physical resources. Those costs matter, and we believe they should be considered alongside the benefits of using the technology.

Beacon is designed to use AI for specific tasks where it can meaningfully reduce educator workload—not simply to add AI to every interaction.

When ordinary software can perform a task effectively, we prefer not to use generative AI unnecessarily.

We do not currently have sufficient information to accurately quantify the complete environmental impact of all third-party infrastructure and AI models that may support Beacon. We will not make environmental claims we cannot substantiate.

As Beacon grows, EmberKeep will continue evaluating how and when AI is used, the technologies and providers supporting our services, and opportunities to reduce unnecessary computation while preserving useful functionality.

We don't believe responsible AI means pretending AI has no costs. It means deciding carefully when those costs are worth paying.

Transparency matters

People should be able to understand when they are interacting with AI-assisted functionality and what role AI plays in the product.

We do not want AI-generated output to masquerade as a professional determination or independently verified information.

As the technology changes, we expect our practices to change too. We will continue evaluating new evidence, provider practices, technical capabilities, and legitimate concerns about AI rather than assuming today's approach will always be the right one.

Our commitments

At EmberKeep, our approach to responsible AI is grounded in four commitments:

People remain responsible.
AI can assist people. It should not quietly replace their judgment.

Information deserves restraint.
We should not collect, expose, or send information to AI systems simply because we technically can.

AI should earn its place.
We use generative AI where it provides meaningful value rather than making AI the default solution to every problem.

Costs count too.
Privacy, accuracy, human impact, environmental impact, and computational resources are real costs. We consider them when deciding whether and how AI belongs in our products.

Responsible technology isn't a finish line. It's an ongoing design decision.

EmberKeep LLC
Arizona, United States
hello@emberkeep.co