How 10AM uses AI.
10AM uses one AI-supported feature in the running customer product: an automated content-safety classification before an uploaded photo can enter a shared album. We also use clearly labelled AI-generated illustrations on some marketing pages. This page explains both uses, their limits, and how to challenge a safety-check mistake.
Last updated: August 5, 2026.
1. The system and its purpose
The service creates a reduced, metadata-free JPEG sample of an uploaded photo and sends it to Google Cloud Vision SafeSearch through Google’s EU endpoint. SafeSearch returns likelihood labels for adult, spoof, medical, violent, and racy content. 10AM applies a fixed rule to those labels to allow or reject the upload for the shared album.
2. What the decision changes
An allowed photo is finalized into the event album. A rejected photo is not added, does not consume the event’s server-side shot allowance, and can be replaced with another shot. If the safety service is unavailable, the upload is not silently accepted; the app keeps the local shot for a retry where supported.
3. Data and records
The classification sample is at most 640 pixels on its longest side, encoded as JPEG, stripped of metadata, and limited to 1 MiB. 10AM stores the allow or reject decision, limited category scores, the version of the 10AM decision policy, the check time, and a SHA-256 digest that binds the result to the finalized image. See the Privacy Policy for processors, retention, and legal bases.
4. Limits and human contact
Automated classification can be wrong and does not understand every context. It is a preventive album-safety gate, not a finding that a person acted unlawfully. If a photo is rejected and you believe that was a mistake, contact support@10am.cam. A person will review the support request; do not email a sensitive image unless support provides a secure way to send it.
5. AI-generated marketing illustrations
Some fictional house-party and wedding scenes on 10AM marketing pages were generated with an AI image tool. They depict illustrative people and events, not real 10AM customers or real event albums, and are labelled as AI-generated where they appear. 10AM does not generate or alter customer event photos.
6. What 10AM does not use
The current customer product does not use AI to generate or alter event photos, identify faces, recognise emotions, categorise people biometrically, socially score people, decide access to essential services, or personalise advertising. The safety classifier runs in the background and does not converse with users.
7. Changes to AI use
We review this notice when the system, provider, purpose, decision rule, or applicable law changes. A materially different AI feature will be assessed before release and disclosed where required.
