Privacy Policy
Last updated: 6 November 2026. See what changed.
Clarify is a social-media moderation platform that helps the teams behind public figures and organisations review and respond to comments on their social pages, and record threats and safety incidents. It is operated by Liam McGrath-Bishop trading as Clarify Socials (ABN 93 729 025 574) ("we", "us"). This policy explains what personal information we collect, how we use it, and the choices available to you. We handle personal information in accordance with the Privacy Act 1988 (Cth) and the Australian Privacy Principles (APPs).
Information we collect
From our customers (account holders)
- Account details: name, email address, and a password (stored hashed). Two-factor authentication is mandatory for every account.
- Sign-in providers: if you choose to sign in with a Microsoft, Google or Apple account, that provider sends us your name, your email address, and whether the provider has verified it. We store the provider's identifier for your account so you can sign in the same way next time. We do not receive your password for that account and we do not read anything else from it.
- Billing details: subscriptions are processed by Stripe. We store your plan, billing status, and Stripe customer reference. We never see or store full card numbers. For an organisation that buys licence seats we also store the email address it gives us for invoices.
- Usage records: actions taken in the app (for example approving a reply, hiding a comment, or deleting an evidence file) are recorded in an Activity Log attributed to the signed-in user.
- Team collaboration content: notes and messages your team writes inside the service (for example review requests, the discussion attached to a draft awaiting approval, account notes, and case notes), stored with the author's name.
- Demo requests: when you ask for a live demo on our website we record your email address, the time, the network address the request came from, and whether you ticked the marketing box. The demo workspace itself contains only made-up data and is deleted a few hours after it is opened. We keep the request record for up to 12 months so we can follow up and to limit abuse of the demo, unless you create an account, in which case it becomes part of your account records.
From someone who invites you
When a customer invites a person to their workspace, or an organisation sends a person a licence or a link to connect their workspace, they give us that person's email address. We use it to send that message on their behalf, to make sure the invitation is used by the person it was meant for, and to show the sender whether it has been used. We do not use it for marketing. Invitations and licence codes that are not used expire (licence codes after 30 days, connection links after 14) and the sender can withdraw them sooner. If you received one you did not expect, you can ignore it, or ask us at support@clarify.net.au to remove your address.
From emails your team forwards
A workspace on a paid plan can have a Clarify email address that its team forwards abusive or threatening emails to. We store each forwarded email as received, including any attachments, together with the address it was forwarded from and the results of the standard checks that show which mail server sent it. We read the forwarded email to find the original sender's name, address, date and subject, and analyse its text in the same way as comments. The original sender is recorded as an account in the customer's workspace. The customer decides what is forwarded and is told not to forward anything confidential or sensitive; we hold what arrives on the customer's behalf, in the same way as comments.
From connected social accounts
When an administrator connects a social account (a Facebook Page, Instagram professional account, Threads profile, TikTok Business Account, X account or Bluesky account), we access, with their authorisation and through each platform's official APIs, content that is already visible to that account, including:
- Posts published by the account, and comments and replies left on those posts;
- The public profile of commenters as exposed by the platform to account admins (such as display name or username, a profile identifier, and where available a profile picture);
- Reaction and engagement metadata on posts and comments, and daily follower counts, where the platform provides them.
Access tokens are encrypted at rest with AES-256-GCM. We do not collect social-platform passwords and we cannot access anything the connected account cannot itself see.
Posts you compose and media you upload
Clarify lets a customer's authorised team compose and schedule posts to publish to their own connected accounts. When you do, we store the post text, your scheduling choices, and any photos or videos you upload, so the post can be published at the chosen time.
- Uploaded photos and videos are stored on our servers for up to 30 days after the post goes live, then permanently deleted. The post itself remains on the social platform, subject to that platform's own policies.
- To publish a post, the photo or video is made available at a web address so the social platform can retrieve it, as those platforms require. The address is a long random name that is not published or listed anywhere, and the file is deleted as described above.
- Posts are only ever published on a person's explicit instruction: Clarify never composes or publishes content on its own.
Incident reports (where available)
A customer's team may record an incident that happened off social media, such as a visit to their office, a phone call or voicemail, a letter or email, a private message, or a vehicle or property incident. These records can contain personal information about people who are not users of any platform and may not be identified, and about the customer's own staff and witnesses:
- Descriptions of a person as observed by the customer's staff: approximate age, height, build, hair, clothing, distinguishing features, a vehicle and its registration, and direction of travel. These are recollections and may be inaccurate.
- A name, phone number, or email address where the person gave one or it is known.
- Files the customer uploads as evidence: CCTV stills and video, photographs, voicemail audio, scanned letters, and screenshots of messages. Each file is stored with a content hash and upload time so a recipient can check it is unchanged since upload.
- The names of staff and witnesses present and, where the customer records it, any injury or impact on them. Health-related detail in an incident report is sensitive information, and we ask customers to record it only with that person's agreement.
- Police reference numbers and a record of what was reported to whom.
The customer collects this information and enters it into Clarify; we hold it on their behalf. The customer is responsible for having a lawful basis to collect it, including under the surveillance, recording, and workplace laws of their state or territory, and for recording what was observed rather than guesses about who a person is or sensitive details that are not relevant to safety. Statements are versioned, so an amendment creates a new version with its author and time and earlier versions are kept.
From an optional Microsoft connection (diary and media library)
A customer may connect their own Microsoft account so that Clarify can show their office diary beside their posting schedule and store event photos and video in their own OneDrive or SharePoint library. Where connected, we read calendar events (titles, times, locations, and descriptions) through the Microsoft Graph API, and read and write media files in the folder the customer chose. The Microsoft access tokens involved are encrypted at rest, and the connection can be removed at any time in Settings.
Media monitoring (Pro and Agency plans)
Where a customer enables media listening, we search publicly available news sources (Google News and the GDELT project, plus any specific outlets the customer chooses) for coverage that mentions them, fetch the text of matching articles from the publisher's website, and store a summary, a tone assessment, and a link to the original story.
Digital card visitors
When a member of the public views one of a customer's public digital cards, we record basic analytics (page views, link clicks, and save-to-contacts downloads) so the card's owner can see how it is performing. These analytics are counts tied to the card, not profiles of the visitor.
How we use information
- To present comments to the customer's authorised team for human review;
- To classify comments automatically (for example flagging abuse, spam, or suspected coordinated activity) and to draft suggested replies for human approval;
- To hide comments automatically, only where the customer's team has set a standing instruction: auto-hiding (off by default), which hides comments the classifier places with high confidence into categories the team selected, and muting, which hides new comments from a specific account the team chose to mute. Every automatic hide is recorded in the Activity Log, queued for human review, and reversible with one click; hidden comments remain visible to the commenter and their friends under the platform's normal hiding behaviour, and nothing is deleted;
- To help teams recognise repeated or coordinated harmful activity: where an account's comments are repeatedly flagged, we may build and retain a profile of that commenter, including an automated suspicion score, behavioural signals, threat or suspected-fake-account indicators, and membership of suspected coordination groups. These profiles are automated inputs to human review, not automated decisions: any resulting action (hiding, reporting, evidence files) is taken by a person;
- To monitor public news coverage for mentions of the customer, where they enable media listening;
- To publish posts the customer's team composes and schedules, to that team's own connected accounts, at the time they choose;
- To build case files, incident reports, and evidence files (for example for reports to the eSafety Commissioner or the police) at the customer's direction. Clarify summarises an incident statement into the case report but never writes or changes the statement itself;
- To operate, secure, and bill for the service, and to send service email such as account confirmation, notifications, and case reminders;
- To send our customers (account holders) marketing and product update communications about Clarify and related products and services, where they opted in (for example by ticking the optional box at sign-up or when requesting a demo) or where the law otherwise allows, as described in our Terms of Service. We record when an account opted in. Every such message identifies the sender and includes a way to unsubscribe, and you can also opt out by emailing us at the address below. We never send marketing to commenters or to anyone else whose information reaches us through a connected social account or an incident report.
Automated decisions
Clarify uses computer programs, including large language models, to make assessments about personal information. This section describes each one, what it can affect, and the person who stays in control.
- Comment classification. Every synced comment is sorted into categories (for example legitimate, abusive, spam, threat, suspected coordination). A classification changes how the comment is shown to the customer's team and whether they are alerted to it. It does not by itself hide, delete, report, or reply to anything.
- Auto-hiding and muting. Only where a customer's team has switched it on, a high-confidence classification in a category they chose hides the comment on the platform pending review, and new comments from an account the team has muted are hidden while the mute lasts. This is the one automated action that can affect a commenter without a person acting first. It is reversible with one click, every instance is logged and queued for review, and a commenter who believes a hide was a mistake can contact the page's team or us.
- Commenter profiles. Repeatedly flagged accounts receive an automated suspicion score, behavioural signals, and suggested membership of coordination groups. These inform a person's review and are never acted on automatically.
- Incident matching (where available). Where several workspaces have granted the same organisation access, the service may suggest that incidents they recorded involve the same person, by comparing the names, phone numbers, email addresses, vehicle registrations, and descriptions those workspaces entered, and nothing else. A suggestion is an indication only. It is shown to the organisation's reviewers, who decide whether it is right, and the workspaces do not see each other's cases unless the organisation shares them.
No automated assessment in Clarify decides whether anyone is reported to a platform, the police, or a regulator, and no automated assessment is used for any purpose other than the customer's moderation, safety, and record-keeping.
Who we share information with
We do not sell personal information. We share it with the processors that run the service, listed below, and, for account holders' contact details only, with our associated entities (businesses we own or control, or that are under common ownership or control with us) so they can send the marketing and update communications described above. Those entities must handle that information in line with this policy and honour any opt-out.
A customer may also connect their workspace to an organisation (for example a party head office or a peak body), either by accepting a link from it or by taking a licence seat the organisation pays for. The connection is an access grant the customer approves before it starts and can withdraw at any time. While it is in place, the organisation's reviewers can see the following, read-only:
- summary figures for the workspace: how many posts and comments it has had, how many comments were classified as harmful and in which categories (for example threats, abuse and misinformation), and how its posts performed;
- the workspace's public posts that drew the most reaction or the most harmful comments: the caption, platform, link and counts;
- accounts that have commented on more than one connected workspace: the account's public name and platform, how many of the organisation's workspaces it has commented on and how often, and whether any of them flagged it as a threat;
- anything the customer's team chooses to escalate to the organisation, such as a case and its evidence, and the messages the two exchange about it;
- where available, the incident-matching suggestions described under Automated decisions.
The organisation does not see the wording of individual comments, the team's notes, or case files unless the customer's team escalates them, and it can never reply, hide, publish, or change anything in the workspace. Where an organisation pays for a workspace's seat it also sees the workspace's name, its plan, the date the seat was taken and the owner's email address, so it can manage its licences. It does not see the workspace's own billing history or payment details. Where the customer also turns on sharing with peer workspaces in that organisation, those peers can see that the customer tracks a given account and whether it is flagged as a threat, and nothing more: comment content and case files never cross between workspaces unless the organisation shares them.
- Meta Platforms (Facebook, Instagram, Threads), TikTok, X Corp. (X, formerly Twitter) and Bluesky Social, PBC (Bluesky): we read and, on human approval, write content to a customer's own connected accounts through each platform's official API, subject to that platform's terms;
- Microsoft, Google and Apple (sign-in): where you choose to sign in with one of these providers, your sign-in request goes to that provider and it returns your name and email address to us;
- Anthropic: content is sent to the Claude API, the language model behind Clarify's automated analysis and drafting: comment text and post captions for classification, plus, depending on the features used, post drafts and topics, diary event details, the customer's stored writing samples, fetched news-article text, and the text of incident statements and case notes for summarising. When drafting, the model may also run web searches through Anthropic to verify facts. Under Anthropic's commercial terms, API inputs and outputs are not used to train their models. Some customers supply their own Anthropic API key instead (stored encrypted): Agency customers, and customers on any plan whose workspace is linked to an organisation. Those requests then run under that customer's own Anthropic account and terms. An organisation can also supply a key for the workspaces linked to it. Where such a workspace has no key of its own, its requests run under the organisation's Anthropic account and terms, the organisation pays Anthropic for that usage, and the workspace's settings show whose key is in use. The content sent is the same whichever key is used. We do not give the organisation a copy of it, and what Anthropic makes available to an account holder about requests made with its key is governed by Anthropic's terms;
- Microsoft (Microsoft Graph): where a customer connects their diary or media library, we read their calendar events and read/write media in their own OneDrive or SharePoint, under the permissions they granted;
- Google News and GDELT: where media listening is enabled, search queries naming the customer are sent to these public news indexes, and article text is fetched from the publishers they link to;
- Google (Google Ads): our public marketing pages and the sign-up form load the Google Ads tag so we can measure which adverts lead to sign-ups. It receives the address of the public page viewed, a random advertising identifier, and the fact that a sign-up was completed. It does not run on any signed-in page, and no customer data, comment, or workspace content is ever sent to it. Details and opt-out are in our Cookie Policy;
- Cloudflare (Turnstile): the live demo request form carries a bot check run by Cloudflare. When you open that form, Cloudflare receives your browser's network address and the technical signals it uses to tell a person from a script, under Cloudflare's privacy policy. It runs on that one form only, never on any signed-in page, and nothing you type into the form is sent to Cloudflare;
- Google (Google Fonts): every page, including signed-in pages, loads the Inter typeface from Google's font servers. That request carries your browser's network address and the page's address, as any web request does, and nothing else. It sets no cookies;
- Stripe: subscription billing;
- Resend: transactional email delivery;
- Microsoft Azure: hosting and data storage in the Australia East region (Sydney, Australia). The database has no public endpoint and is reachable only from inside our private network.
We may also disclose information where required by law, as described in our data request policy. Some of the processors listed above operate outside Australia and may hold personal information there: Anthropic, Stripe, Resend, Google, Apple, Meta, X Corp. and Bluesky in the United States, and TikTok in Singapore and the United States. Microsoft's sign-in service may process the sign-in request outside Australia. We take reasonable steps to ensure each of them handles personal information consistently with the APPs, and the content of your workspace (comments, posts, cases, and files) is stored only in Australia.
Storage and security
- Data is hosted on Microsoft Azure in Australia, encrypted in transit (TLS) and at rest;
- Files uploaded to an incident report or case are stored in a private container in the same Australian storage account. Nothing in it is reachable by a web address: files are streamed only to signed-in members of the owning workspace, and to the organisation it has escalated the case to;
- All connected-account access tokens (the social platforms and any Microsoft connection) are additionally encrypted at the application layer;
- Secrets are held in a managed vault, the database is not reachable from the public internet, and access by our customers' staff requires mandatory two-factor authentication;
- Customer data is segregated per organisation (tenant) with enforced isolation.
If a data breach is likely to result in serious harm to anyone whose information we hold, we will notify the affected customers and the Office of the Australian Information Commissioner as the Notifiable Data Breaches scheme requires, and work with the customer to notify the people affected.
Cookies
Inside the service we set only the cookies needed to keep you signed in and to keep it secure. On our public marketing pages and the sign-up form, the Google Ads tag sets advertising measurement cookies so we can tell which adverts led to a sign-up; it never runs on a signed-in page. We do not use analytics, session recording or social media tracking pixels anywhere. Each cookie, what it does, how long it lasts, and how to opt out of the advertising ones is listed in our Cookie Policy.
Platform operator access
A small number of authorised Clarify personnel can access customer environments where necessary to operate, support, and secure the service. For example diagnosing a failed page sync, investigating suspected abuse of the platform, or assisting with an account issue. In-app operator accounts are subject to the same mandatory two-factor authentication as every other account, and actions taken in the application are recorded in the Activity Log. Direct access to the underlying infrastructure and database is limited to authorised engineers over restricted network paths and is read-only except where maintenance requires changes. Operator access is not used to read customer content for any other purpose.
Retention and deletion
We keep synced content and account information while the relevant account connection and customer account are active. When a connected account is disconnected or a customer account is closed, the associated data is deleted on request, see our data deletion instructions for the exact steps. Backup copies are removed in line with our backup retention cycle.
Photos and videos uploaded for a composed post are retained for up to 30 days after the post goes live and then permanently deleted from our servers automatically; a draft's media is removed when the draft is deleted. The published post remains on the social platform until removed there.
Forwarded emails and their attachments are kept on the same terms as comments. One exception: an email from an address the workspace has not approved is held for the workspace's owners and deleted automatically after seven days unless an owner approves the sender.
Incident reports, their earlier versions, and files uploaded to a case are kept until the customer's team deletes them or the workspace is deleted. They are never removed by automatic clean-up. A workspace administrator can delete a file, a statement, or a whole case at any time; each deletion is recorded in the Activity Log, and a deleted file can be recovered from storage for 14 days, after which it is gone.
We may retain specific content beyond a deletion request where it forms part of an active safety case or evidence file (for example documentation of threats, abuse, or suspected coordinated harassment), has been provided to or prepared for law enforcement or a regulator, or where retention is required by law or reasonably necessary to establish, exercise, or defend a legal claim. The data deletion instructions describe how these grounds are applied and communicated.
Your rights
You may request access to, correction of, or deletion of personal information we hold about you, whether you are a Clarify customer, a person whose comment on a connected account was processed by the service, or a person described in an incident report or case. Contact us at support@clarify.net.au and we will respond within 30 days. Where the information is in a versioned incident statement, a correction is recorded as a new version alongside the original rather than by overwriting it. Deletion requests may be declined or deferred on the limited safety-and-evidence grounds described in our data deletion instructions. If you are unsatisfied with our response you may complain to the Office of the Australian Information Commissioner (oaic.gov.au).
Contact
Clarify: support@clarify.net.au
