Privacy Policy

1. Information we process

• Candidate information - names, contact details, résumés/CVs, work history, qualifications, availability, salary expectations, recruiter notes, and consent records.
• Client and hiring-manager information - organisation details, contact names, job titles, work email addresses and phone numbers, and terms of business, held in the connected CRM or entered by the Agency.
• Placement and fee information - for each placement, the candidate and job, the salary or pay rate, the fee and its percentage, the start date, and its status.
• Business-development leads - organisations that may need to hire, built from the Agency's CRM history and from roles organisations advertise publicly on their own careers pages and public applicant-tracking boards (organisation and role details only, read only where the site's robots.txt allows it).
• Pronoun - a candidate's pronoun (His or Her), set by a recruiter or suggested automatically, used for gendered wording in generated emails and submission packs.
• Behavioural profiles - where an Agency switches behavioural assessment on, a candidate or client contact may complete a working-preferences questionnaire. We hold the consent record, their answers and a derived behavioural profile with a headline summary. It describes working preferences, not ability, health or personality type. A person can decline the assessment or stop part-way. Unless the Agency has opted out, a de-identified record of each placement with a recorded outcome is also added to a pool shared across Agencies: behavioural bands for the placed person, the role and, where recorded, the hiring manager and team; how well they matched; the sector and role family; whether the placement worked out (probation, retention, promotion, and a broad reason if it ended) with broad satisfaction bands; and the quarter. It carries no names, identifiers, free text, salaries or exact dates, and a pattern is shown only where at least 20 placements from at least 3 Agencies support it. An Agency can opt out in Settings, which deletes everything it contributed.
• AI-generated assessments - fit scores, the selection criteria a candidate meets or misses, shortlist rankings and the judgements and rationales behind them, and comparisons of a candidate with a role.
• Interview records - interview plans, schedules, summaries, key points, outcomes and the notes recorded with them.
• Learned preferences and remembered facts - how a recruiter likes things written, what makes a good match for them, saved "plays", and facts a recruiter asks TalentDesk to remember about a candidate, client or job (including a note a recruiter adds when moving a candidate between stages). These are held against the recruiter's account and, for facts, the person or job they concern.
• Phone app Ask conversations - the questions a recruiter asks in the TalentDesk phone app's Ask, the answers, and the cards shown with them, which can include candidate and client details. Each conversation is kept for 1 hour to 7 days after its last message, as the recruiter chooses.
• Files added to the phone app's Ask - documents and photos a recruiter adds to a question in the phone app's Ask (PDF, Word, JPEG, PNG or plain text, up to 20 MB each and five to a question), which can include candidate and client details. They are kept in TalentDesk's private file storage.
• Problem reports - what a recruiter tells us went wrong when they report a problem in TalentDesk, the page they were on and the error reference it showed, which can mention candidate or client details, and the team's questions and the recruiter's answers about it. No screenshot is taken.
• User (recruiter) information - name, work email, job title, profile details, and authentication identifiers.
• Communications and calendar data - emails, drafts, calendar events, and interview/meeting details (including optional, consent-gated interview recordings and transcripts) accessed through Microsoft 365 with your authorisation.
• Usage and technical data - log data, device and browser information, and audit records of actions taken in the Service.

2. Where it comes from

We receive information from: connected recruitment CRM systems (Bullhorn and JobAdder) that your Agency authorises; Microsoft 365 (Outlook, Calendar, Teams) when a user connects their account; Veritone Hire (Broadbean), when an Agency connects it: the applications and CVs people send through the job boards the Agency advertises on, and the candidates a recruiter imports from job-board searches; information that Agencies and recruiters enter directly; the public careers pages and applicant-tracking boards of organisations, for business-development leads; and automatically as you use the Service.

3. How we use it

We use information to provide and operate the Service - including syncing jobs and candidates, shortlisting, generating submission packs, scheduling and (where enabled) recording interviews, drafting and sending communications you initiate, maintaining security and audit logs, and improving and supporting the Service.

4. AI-assisted features

The Service uses artificial-intelligence models to help draft, summarise, extract, and rank information. Content you process (which may include candidate and role information and message text) is sent to third-party AI providers via the Vercel AI Gateway. Every AI model we use processes your content under zero data retention terms, which are the Vercel AI Gateway's agreements with the providers: the provider uses it to produce its answer and does not store it beyond that, apart from narrow exceptions some providers' terms allow, such as where the law requires it, and providers under these terms may not train their models on it. This covers every model we use, including TypeSafe AI's Jev evaluation model and the voice model behind the phone app's live voice mode. Every request our servers send to an AI model requires zero data retention. A request that can't be served with zero retention is refused, and a nightly test confirms it. The phone app's live voice mode is the one AI connection our servers do not send: when a member starts a voice conversation, the phone streams their voice through the Vercel AI Gateway to xAI's Grok voice model, which holds the conversation and asks TalentDesk for anything about your records. That connection cannot carry our zero retention requirement, so it relies on the Gateway's zero data retention agreement with xAI: a voice conversation starts only while the Gateway lists the model as served with zero data retention, we check that listing before every conversation and every night, and we do not record the conversation. AI processing takes place outside Australia, for example in the United States. Providers currently include Anthropic, TypeSafe AI and, for the phone app's live voice mode, xAI, and the cloud platforms that host models for the Gateway: Amazon Web Services, Microsoft Azure (which serves an OpenAI embedding model we use to match people to roles) and Google Cloud. AI output can be inaccurate and should be reviewed before it is relied on or sent. If you connect your own AI assistant to TalentDesk (for example through our Model Context Protocol endpoint), the data you request through it is sent to that assistant's provider under that provider's terms. By default an assistant you connect sees people's names as a first name and surname initial, with contact details masked. An Agency owner or administrator can approve fuller data for a specific assistant, and can revoke that approval at any time; an Agency owner can also switch off the assistant's access altogether. Each file a recruiter adds to the phone app's Ask is checked for viruses by TalentDesk's own scanner, which keeps nothing and answers only clean, infected or failed; a file found infected is deleted. The text of a PDF or Word file is read on TalentDesk's servers; a scanned PDF's first pages and a photo are transcribed by an AI model through the Vercel AI Gateway, and the Ask's model reads the file's text the same way, both with zero data retention.

5. How we share it

We share information with: the connected CRM (to write back submissions and placements you authorise, and to file a document from the phone app's Ask on the record a recruiter chooses, once they confirm it with a press and hold; the Agency's own CRM retention then applies to that copy); Microsoft (to send email and create calendar events you initiate); Veritone Hire (Broadbean), when an Agency connects it (to publish the job advertisements recruiters post and take them down; it receives the advertisement and the recruiter's Broadbean username, not candidate records); and our service providers who host and run the Service - including Supabase (database and authentication, hosted in Sydney, Australia); Vercel (application hosting, page-speed measurement, and the AI Gateway that connects us to the AI providers described above); Upstash (a short-lived cache, hosted in Sydney, Australia); Inngest (runs background jobs, United States; it receives record identifiers, counts and status, some job and client details, and shortlist results linked to record identifiers, but not candidate contact details); Resend (sends emails from our sign-in service, Japan); Cloudflare (domain name services, forwarding of email sent to our addresses, the bot check on our sign-in and sign-up pages, and access control for our staff console); Google (Firebase Cloud Messaging) and Apple (Apple Push Notification service), which deliver the phone app's notifications: they receive the device's notification token and a short fixed message with a record identifier, not names or contact details; and the AI providers described above. We do not sell personal information. We may also disclose information where required by law.

6. Where it is stored

Our database, sign-in service and file storage are hosted in Sydney, Australia (Supabase, on Amazon Web Services), and the application runs in Sydney (Vercel). Some information is disclosed to service providers outside Australia: AI providers reached through the Vercel AI Gateway (outside Australia, for example in the United States, with zero data retention), Inngest (United States) and Resend (Japan). Google and Apple, which deliver the phone app's notifications, and Cloudflare, which forwards email sent to our addresses and runs the sign-in bot check, may also process information outside Australia. When an Agency connects it, Veritone Hire (Broadbean) also receives information and may be outside Australia. If you connect your own AI assistant, the information you request through it goes to that assistant's provider.

7. Security

We protect information with encryption in transit (HTTPS) and at rest (AES-256, by our database host), a second layer of TalentDesk encryption on the credentials used to connect CRMs and Microsoft 365, separation of each Agency's records enforced by the database, tested before every release and checked in production every day, sign-in through Microsoft 365 with access only by invitation or Agency approval, role-based permissions, a separate audited console for our staff, and audit logging. Our Security page has more detail. No method of transmission or storage is completely secure, and we cannot guarantee absolute security.

8. Retention and deletion

We keep information in the Service for as long as we need it to provide and develop the Service, and as required by law. We do not delete candidate, client, job, placement, behavioural, interview or audit records on a timer, so there is no fixed number of months or years after which they are deleted. A few things are deleted automatically:

• Platform diagnostic logs and the error records drawn from them, which can incidentally contain a name or email address, after 30 days.
• Summaries of recurring errors, which hold no personal information, 180 days after the error was last seen.
• Facts a recruiter has asked TalentDesk to remember expire unless they are learned again: a fact about a client after 18 months, a fact about a candidate or a job after 12 months, and a note after 6 months (a note that has never been used is deleted by 9 months at the latest). A fact a recruiter forgets can be brought back for 7 days; after that its text is erased and only a marker that it was forgotten is kept, so it is not learned again. Taught preferences and outcomes do not expire.
• Writing samples taken from a recruiter's sent email are deleted when that recruiter has not allowed TalentDesk to read their mailbox, or withdraws that permission. A sample a recruiter removes is erased after 30 days.
• Records of the sync with an Agency's CRM (the result of each run and the health checks on it), after 90 days.
• A recruiter's problem reports, with the team's questions and their answers, 180 days after the report was sent.
• The confirmation step for an action requested through a connected AI assistant: what the action would do is erased when it is confirmed, or by a sweep that runs every 15 minutes once its confirmation window closes, and the record is deleted 7 days after it expires.
• A recruiter's current chat with the TalentDesk phone app's Ask is kept for 1 hour, 24 hours or 7 days after its last message, as they choose (24 hours by default), then deleted; they can delete it sooner.
• Files added to the phone app's Ask: a file not sent with a question is deleted 1 hour after it is added, and its record 3 hours after. A file sent with a question is kept with that chat for as long as the chat is kept, and at most 30 days after it was first sent. Removing it from the question, starting a new chat, closing the Ask with files not yet sent, the chat's own expiry and a candidate's suppression delete it sooner.
• Files that arrive with an application from Veritone Hire (Broadbean), when an Agency connects it: once the application has been imported, and after 7 days at the latest. The copy of the applicant's CV shown in a recruiter's Inbox is deleted once the application is decided, or after 30 days unless it is still being added to the CRM.
• A CV a recruiter uploads for a candidate and never saves to their record: after 24 hours. One whose save to the CRM is still under way, or could not be confirmed, is kept until that save is settled. When more than 900 Agencies have one due in the same hour, the rest wait for later hours, the Agencies with the oldest uploads first.
• Sign-in records for connected AI assistants: an unfinished sign-in request or an unused code 1 day after it lapses, the access tokens of a connection that has ended 30 days after it ended, and an assistant's registration that nobody approved 90 days after it was made.
• Records of the notifications sent to the TalentDesk phone app: after 30 days.

Otherwise records stay in the Service. A request to remove one is handled as described below.

When an Agency asks us to remove a candidate, the candidate is hidden from that Agency's workspace and board and can be restored; the underlying record is kept, and it keeps syncing from the Agency's CRM. Erasure is not a self-service control: the Agency raises the request, or you contact us, and a TalentDesk administrator handles it by hand. A tool that deletes or de-identifies every copy of a record is not built yet, so we may not be able to remove every copy on request; when we cannot, we tell you what we have kept and why. Answers to a behavioural assessment, and the profile derived from them, are covered by the same request.

Candidate records are also held in the Agency's connected CRM. Our sync adds and updates records; it does not delete them, so a record deleted in the Agency's CRM is not automatically removed from the Service. Deletion and erasure requests are handled as described above.

9. Your choices and rights

You can ask us for access to, or correction of, the personal information we hold about you, or ask us to delete it, by emailing support@talentdesk.com.au. You can also contact the Agency you dealt with. We will respond within 30 days and tell you what we have done or why we have not. To complain about how we have handled your personal information, email us; if you are not satisfied with our response, you can complain to the Office of the Australian Information Commissioner (OAIC) at www.oaic.gov.au. Australian privacy law (the Privacy Act 1988 and the Australian Privacy Principles) applies to our handling of personal information.

10. Children

The Service is intended for use by recruitment professionals and is not directed to children.

11. Changes

We may update this policy from time to time. The "Last updated" date above shows when it last changed; material changes will be reflected here.

12. Contact

Questions or privacy requests: support@talentdesk.com.au.