Legal · Annex

Sub-processors & Annexes

Last updated: 9 September 2026

Forms part of: the O’Care Privacy Policy. This page holds the parts of that notice that change with ordinary engineering and vendor work — the sub-processor register, the AI risk tiers, and the device-permission list.

Notice of changes: we give participating Schools at least 30 days' notice before adding a new sub-processor. Changes to this page do not by themselves change the Privacy Policy, which carries its own "Last updated" date and its own 14-day notice.

1. Sub-processor register

"Region" is where the processing takes place, not where the company is incorporated. The Agreement column records the instrument that governs each party — a signed Data Processing Agreement where we hold one, or the platform or service terms that apply where the relationship is not one we contract individually.

Core service

Sub-processorPurposeData categoriesRegion
Container hosting providerApplication and database hostingAll account and content dataMalaysia (ap-southeast-5)
OC Global Technology AI serviceContent moderation (operated by us, on infrastructure we control)Content submitted for analysis, in memory only — not written to durable storage by this serviceMalaysia
Cloudflare R2 (Cloudflare, Inc.)Media storage — avatars, post and message attachments, voice notesUploaded filesAsia-Pacific
AWS Simple Email Service (Amazon Web Services)Transactional email from the app and the dashboardEmail address, message contentSingapore (ap-southeast-1)

Notifications and messaging

Sub-processorPurposeData categoriesRegion
Expo Push Service (650 Industries, Inc.)Relays push notifications onward to FCM and APNsDevice push token, notification title and bodyUnited States
Firebase Cloud Messaging (Google LLC)Push notification delivery to Android devicesDevice push token, notification title and bodyUnited States / EU
Apple Push Notification service (Apple Inc.)Push notification delivery to iOS devicesDevice push token, notification title and bodyUnited States / global
Mocean SMS delivery of SOS alerts to emergency contactsRecipient phone number, alert messageMalaysia

Diagnostics

Sub-processorPurposeData categoriesRegion
Sentry (Functional Software, Inc.) Mobile crash and error reportingCrash reports, stack traces, device and OS identifiers, breadcrumbs, a sample of performance traces. Automatic IP-address and username attachment is disabled.United States

Public website only

The parties below serve the public website at ocare.ocgt.app. None is used by the O’Care mobile app. Analytics, bot protection and contact-form delivery are confined to the public website and receive no account data. Google Fonts is the exception: the administration dashboard loads web fonts from it, including on its sign-in pages and inside the authenticated app, so Google receives the IP address and browser user agent of dashboard users as well as website visitors. No other data reaches it.

Sub-processorPurposeData categoriesRegion
Google Analytics 4 & Google Tag Manager (Google LLC)Aggregate traffic measurement. Google Signals is disabled, no advertising or remarketing feature is enabled, and event-level data is retained 14 months.IP address, user agent, pages viewed, referring sourceUnited States / global
Cloudflare Turnstile (Cloudflare, Inc.)Bot protection on the contact formIP address, browser behavioural signalsGlobal CDN
Google Gmail API (Google LLC)Delivers contact-form submissions to an O’Care mailboxThe enquiry you type, plus your IP address and browser user agentUnited States / global
Google Fonts (Google LLC)Web font deliveryIP address, user agentUnited States / global CDN

The transfer condition relied on for cross-border recipients is a Transfer Impact Assessment together with contractual safeguards — see Privacy Policy, Section 9.

Agreement column. Every row is marked to be confirmed until the executed instrument has been located and its counterparty entity checked; see the review note.

2. AI risk tiers

Privacy Policy, Section 6 describes the automated content moderation. This annex sets out the risk levels the analysis assigns and what happens at each, because those thresholds are tuned as the models improve.

Risk levelWhat happens
LowNo alert is raised. For a diary entry, post or chat message the verdict is recorded against the content and contributes to the aggregated daily wellbeing summary shown to authorised staff and, where applicable, to parents. For status/story content, which is not stored after it expires, nothing is recorded.
MediumAs Low: no individual alert.
HighAn alert is created for authorised school coordinators and administrators. Where an alert about the same person is still unresolved, the new finding is added to it rather than raising a second one. Posts are withheld from other users and marked for review; chat messages are hidden from the conversation, and where the hidden message was the most recent in that conversation its preview is replaced with a placeholder. The author is notified that their content was flagged.
CriticalAs High. In addition, where the content is the author's own diary entry, the app shows crisis resources to the author — Talian Kasih 15999, Befrienders Malaysia +60 3-7956 8145, and a link to their school counsellor — before any other action is taken.

Withholding content at High or Critical is an automated step, taken to prevent harm while a person assesses it. It is not a penalty on the account. Diary entries are visible only to their author and to authorised reviewers, so they are not withheld. Formal cases, disciplinary outcomes and referrals to law enforcement are always decided by a person.

You may contest any AI assessment using the "Tell us this is wrong" control in the app, and request human review under Privacy Policy, Section 11.

Alert records hold metadata only — risk level, reason, content type, a pointer to the source record where the content type carries one, and a hash of the analysed text. They do not hold the raw content. Resolved alerts are purged after 90 days; unresolved alerts are kept until reviewed.

3. Device permissions

O’Care requests the permissions below on a strict need-to-use basis. Each can be revoked from your operating system's settings at any time; revoking one disables the feature that needs it and nothing else.

PermissionWhyWhen asked
CameraProfile photo, posts, chat attachmentsWhen you tap the camera button
Photo LibraryUpload media to posts, chat, profileWhen you tap the gallery button
MicrophoneVoice notes in chatWhen you start a voice recording
NotificationsDeliver messages, SOS responses, and admin alertsAt first launch
LocationAttach your location to an SOS alert so responders can find youWhen you trigger SOS
ContactsImport an emergency contact without retyping it. The picker runs on your device, and the emergency-contact list you build is stored on your device. A contact’s name and number reach our servers only when you trigger an SOS and we notify them, or where you supplied one emergency contact during onboarding — that one is stored on your account. When you tap to import a contact