Trust Center

Where a model is in the loop.

The places LiftLine AI uses a model, what each receives, what none may decide, and an honest account of forecast accuracy.

Several parts of LiftLine involve a model, and they are not the same system.

Where a model is used

Counting people in lift lines

A vision model on our own computer reads a frame from a public resort webcam and counts the people in the lift line. Each frame is checked against a reference image first, to confirm the camera is pointed at the lift we think it is. The output is a head count, a confidence and a timestamp. It does not identify anyone: no face recognition, no matching people across frames.

The forecast

Since August 31, 2026 the forecast is a simulation: simulated riders move over a graph of the mountain in 15-second steps, lift queues form from each lift's capacity, and webcam head counts are fed in through the day to keep the simulated crowd in line with what the cameras see. Published forecasts are per lift and per hour. Nothing about you goes into it.

Liftie, the assistant

When you talk to Liftie, your message and the rest of that conversation go to our own server. The server adds your riding profile and the notes Liftie keeps about you, and your name if you turn on Use my name, and sends the conversation to Anthropic's Claude, which writes the answer. If you ask where you are, the app sends the names of the nearest lifts, not your coordinates. Our server does not save your messages; Anthropic does not train on them and deletes them within 30 days.

Short lines on your cards

When your iPhone supports Apple Intelligence, Apple's on-device model writes the short line on a card, and the explanation of a conditions report, on the phone itself. With Max, a personal line your iPhone cannot write is written on our own server instead, from the same profile and notes as Liftie.

Reading public resort news

We sometimes use outside AI tools, including Anthropic's Claude, to read public resort news. They never receive anything about you.

What a model is allowed to decide

A model decides
How many people it thinks are in a picture, and, inside the simulation, what a wait will be.
A model does not decide
Whether you get access to anything, what you are charged, whether a report of yours is kept, or anything else with a consequence attached. There is no automated decision-making about you in this product.

How accurate the forecast is

Scores by resort

Each resort we forecast, whether its forecast has been scored, and how it did against two simple guesses.

Live figures load in your browser.

Yesterday, lift by lift

For each lift a camera watches: the forecast we set before the resort opened, the wait the app showed through the day, and the wait the camera observed.

Live figures load in your browser.

A forecast can only be checked where a camera watches the lift line, so scoring will cover the lifts with a camera, and we will say how much of the forecast that is. A forecast is an estimate for planning, not a promise about the minute you reach the lift.

We retired the previous forecast on August 31, 2026, and its scores with it. They described a system that no longer runs, so they are no longer published.

Training

  • We do not train any model on your data.
  • We train two models: a terrain classifier on satellite and lidar imagery of mountains, and the crowd counter on public webcam frames.
  • Anthropic answers Liftie's chat under terms that forbid training on it and delete it within 30 days.

Turning it off

Liftie is optional. You can use lift status, conditions and forecasts without ever opening it, and you can turn it off in Settings › Liftie AI. Controls and retention are in the policy at https://liftlineai.app/privacy/.

The public pages of this site are written by people.

Last reviewed October 3, 2026.