1
Sandboxed tools
Its Python and file tools run in a sandbox. Those tools have no network access and cannot reach anything outside the AI’s own folder — and nothing on the founder’s computer or phone.
Safety
Built to stay inside the lines.
Three safeguards are permanent rules for Ultrec. They are not settings we switch off when it is inconvenient.
2
Hidden keys
Access keys stay hidden. The AI uses the services it needs without being able to see, copy, or leak the keys behind them.
3
Daily caps
It stops when it reaches its daily usage cap. It does not run forever, and it does not raise its own limit.
Oversight
Separate checkers keep an eye out.
Independent watchers review how the AI behaves and look for signs that a safeguard has been tampered with. If something looks off, a person takes a look. It is routine housekeeping, the way any careful team double-checks its work.
What this means for you
Plain and honest.
- AI can be wrong. We say so, and we don’t publish accuracy numbers we can’t back up.
- Our helpers can’t reach your devices through their tools.
- Limits are part of the design, not a bug to work around.
- Questions about safety? Write to us — we read every note.