We’re constantly asked about how we handle data that’s exposed to the Sieve platform. The simplest way to answer this question is that Sieve is able to function without having to store any sensitive raw data on our servers.

Sieve’s Data Requirements

Sieve needs to following to function:

• Raw videos OR secure URL to raw videos that you host
• Any additional metadata that you’d like to have searchable within Sieve

Sieve is able to do all of its processing without having to ever permanently store any of your sensitive information on Sieve servers. A user can host their videos in cloud storage services like S3, GCP buckets, or some other similar service - which they can use to provide us a URL to the video, rather than the raw video itself. This reduces the amount of data sent to Sieve and makes processing much faster because there’s no delay of having to copy data to a Sieve server.

We only access raw data for the following purposes:

• Data processing: We need to run our proprietary computer-vision models and filters over your videos to efficiently generate metadata. Your raw data is loaded into memory while our models process it. However, your raw data is never permanently stored anywhere.
• Visualization: If you decide to use our dashboard to visualize your queries, we access the raw data URLs on our web client to render the images in your browser.

Any data URLs sent to Sieve can also be access-controlled such that can only be seen for a certain amount of time and only by Sieve. Sieve includes its own authentication headers, so users can expose data only to Sieve.

• On AWS
• On GCP
• On Azure

Typically, users give Sieve 1 hour of time-limited access for any operations it needs to perform, after which it has no access to the raw data. Instead, it only hosts the metadata generated from the data processing.