Trust through transparency
Why hu asks first, shows its work and never acts on its own.
The most capable AI system in the world is useless if people don’t trust it. And trust isn’t something you can build through marketing language or privacy policies alone. It has to be embedded in the design of the system itself.
At huSpace, we’ve developed a set of principles for building AI that people can actually trust — not because we tell them to, but because the system behaves in ways that earn trust over time.
Predictability over magic
There’s a temptation in AI to optimize for impressive demos — moments where the system does something unexpectedly clever that makes people say “wow.” But in daily use, unpredictability is the enemy of trust.
We design hu to be predictable. When you ask it to do something, you should be able to anticipate roughly how it will respond. When it takes an action, you should understand why. This doesn’t mean the system is limited — it means it behaves consistently within a well-understood scope.
Explicit permission, not inferred consent
Many AI systems operate on the assumption that if you’re using the product, you’ve consented to whatever the product does. We take a different approach: actions that affect the outside world require explicit permission.
Permission architecture
- Safe actionsExecute right away
- Set reminders
- Add to lists
- Answer questions
- Confirm actionsAsk before acting
- Send messages
- Schedule meetings
- Change your calendar
- Protected actionsRequire authentication
- Payments
- Account changes
- Sensitive data
This means hu will ask before sending a message on your behalf, before acting on your calendar (a connection that’s next), before accessing a new source of information. Over time, you’ll be able to grant standing permissions for routine actions — but the default is to ask.
Visible memory
For hu to be genuinely helpful, it needs to remember things about you — your preferences, your relationships, your ongoing projects. But memory that operates invisibly is memory that can’t be trusted.
hu’s memory is fully visible: you can see what it knows about you, edit or delete any fact, and open the receipt on any reply to see which facts it used — and which it held back. This transparency is essential for the kind of long-term relationship we’re building.
Product UI · sample dataNo hidden autonomy
Some AI systems are designed to act autonomously — making decisions and taking actions without explicit instruction. We believe this is a mistake for personal AI, at least at this stage of the technology.
hu does what you ask it to do. It may offer suggestions, reminders and observations — but it doesn’t take actions on its own. You remain in control of your digital life, with hu as a capable assistant rather than an autonomous agent. Check-ins, when they come, will follow the same rule: a suggestion, never an action.
Trust is earned incrementally
We don’t expect people to trust hu fully from day one. Trust is built through consistent, predictable behavior over time. Our job is to design a system that earns that trust through every interaction — and never takes it for granted.
