Responsible AI

Seven principles guiding how we develop and deploy AI at AutiPower.

1

Humans Always Make Final Decisions

AI provides suggestions and information. But you decide what's right for your child and family. We refuse to automate important parenting decisions. Every piece of guidance comes with context, limitations, and encouragement to consult professionals.

How we practice this:

  • You decide which suggestions to try
  • You decide when to consult a therapist
  • You decide what's best for your child
  • You consult professionals for anything important
2

AI Provides Information, Not Authority

We never position the AI as an expert, authority figure, or replacement for professionals. AutiPower is informational support, not clinical guidance. Parents should feel empowered by information, not dependent on it.

How we practice this:

  • We say "families often find..." not "you should..."
  • We say "research suggests..." not "clinically proven"
  • We say "consider consulting..." not "you need..."
  • We acknowledge the limits of what AI can understand
3

Transparency Over Automation

We show our work. Instead of giving you a quick answer, we explain how we arrived at it. You can see the reasoning. You can disagree. You can seek professional guidance. Transparency lets you evaluate information critically.

How we practice this:

  • Every response explains the reasoning
  • We disclose what data we're using
  • We show safety logic when escalating
  • We admit when we're uncertain
4

Safety Over Convenience

It's slower to validate every response. It's slower to escalate to professionals. It's slower to be transparent about limitations. But safety is not negotiable. We build time for safety into our processes, even if it means slower responses.

How we practice this:

  • We validate responses before showing them
  • We escalate to 988 and 911 for emergencies
  • We refuse requests for harmful advice
  • We pause rather than guess
5

Understanding Over Compliance

We want you to understand your child better. We want you to understand autism better. We want you to understand yourself better as a parent. Not just follow steps or implement strategies. Deep understanding leads to better decisions than compliance.

How we practice this:

  • We explain the "why" not just the "what"
  • We encourage reflection, not just action
  • We share context and research, not just tips
  • We celebrate understanding, not outcomes
6

Privacy Over Growth

We could grow faster by selling your data or using it for targeted advertising. We won't. Family privacy is sacred. We keep data minimal, share nothing without consent, and design every feature with privacy first. Long-term trust matters more than short-term growth.

How we practice this:

  • We never sell family data
  • We never share with third parties without consent
  • We never use data for advertising
  • We delete data you request deleted
7

Long-Term Trust Over Rapid Expansion

We'd rather serve 1,000 families really well than 1 million families poorly. We'd rather build slowly with community input than rush features to market. Trust is built slowly. Once lost, it's lost forever. We design for long-term credibility.

How we practice this:

  • We prioritize safety over feature speed
  • We involve community in decisions
  • We publicly acknowledge limitations
  • We improve based on feedback

Our Commitment

These are not aspirations. They are commitments we live by every day. If you see us violating any of these principles, please tell us immediately. We want to be held accountable to this standard.

How We Make This Real

These principles require actual work. Here\'s how we implement them.

Safety Layer

  • Multi-layer safety system detects harmful requests
  • Risk classification (LOW/MEDIUM/HIGH/CRISIS)
  • Automatic escalation to 988, Crisis Text Line, 911
  • Response validation before showing to users
  • Continuous improvement based on error reports

Design Layer

  • Every response includes context and limitations
  • Encouragement to consult professionals
  • No guarantees or outcome promises
  • Exploratory language, not prescriptive
  • Clear scope: informational, not medical

Governance Layer

  • Advisory council with parents, educators, advocates
  • Regular safety and ethics reviews
  • Transparent error reporting mechanism
  • Community feedback process
  • Public accountability for decisions

Privacy Layer

  • Minimal data collection (only what's necessary)
  • No data sharing without explicit consent
  • No advertising or profiling
  • User control over all data
  • Data deletion on request
Trust & Safety Info