Google analytics are off.

Privacy Policy

Last updated: August 31, 2026

This Privacy Policy describes how Promptective (“Promptective”, “we”, “us”) collects, uses, discloses, and protects personal information when you visit our website, create an account, or use our workforce AI security products, including our web dashboard, APIs, browser extension, desktop application, command-line tools, and endpoint agents (together, the “Services”).

We build security software and apply the same discipline to personal information: routine records are content-minimised, metadata and cryptographic hashes are preferred over content, and conditional or exceptional processing is identified below.

At a glance

Content-minimised routine records

Prompt and response text can be analysed transiently for threat detection. Default hosted interaction records exclude the text and store bounded metadata with SHA-256 hashes. A legacy raw-content switch is restricted to isolated development or testing with non-sensitive data and is prohibited for production and customer data.

Optional analytics with regional controls

For visitors detected in Australia, Google Tag Manager and configured analytics tags are on by default and can be turned off at any time. They remain off until allowed in regions that require prior opt-in and when location is unknown or has not been assessed for opt-out. The setting lasts for 180 days and never grants advertising consent.

Metadata-first telemetry

Routine product security events record surfaces, verdicts, timestamps, and identifiers. Raw prompt, response, and source content are deliberately excluded from routine telemetry.

Your controls

Organisation administrators control protection modes and policies and can record enterprise retention limits. Those settings do not by themselves prove that deletion has run. You can request access, correction, or deletion at any time.

01

Who we are and what this policy covers

Promptective provides workforce AI security across browser, endpoint, gateway, and code-assistant surfaces, including supported MCP-aware code-assistant hooks. This policy covers the public website at promptective.ai, the web dashboard and APIs, and installed clients such as the browser extension, desktop application, command-line tools, and endpoint agents.

You can raise any question or request under this policy with privacy@promptective.ai or through the privacy contact form linked at the end of this policy.

02

Controller and processor roles

For personal information collected through the website, account registration, contact forms, and billing, Promptective is the data controller.

For routine security telemetry generated when an organisation deploys the Services, we process personal information on that organisation’s behalf and under its instructions. The organisation’s administrators decide protection modes and policies and may set retention-policy limits, subject to the deployed service and customer agreement. If you use the Services through your employer or another organisation, please direct privacy requests about that data to your administrator; we support organisations in responding.

03

Account and organisation data

When you register, or are invited to an organisation, we process:

  • Your name, email address, and email verification status;
  • Authentication records: passwords are stored in salted, hashed form; session and verification records contain opaque token values or identifiers and expiry information; Promptective API keys, installation credentials, and enrolment tokens are stored server-side as cryptographic hashes while raw credentials stay outside those records;
  • An optional profile image, and organisation details such as name, logo, membership roles, and invitations (including invitee email addresses);
  • Session records used to keep you signed in and secure your account, including a session identifier, expiry, IP address, and browser user-agent string;
  • Rate-limit counters that protect authentication and account endpoints against abuse.

04

Website enquiries

When you submit a contact, demo, or partner enquiry, we collect the details you choose to provide, which can include your name, work email address, company, role, phone number, region, website, team or client size, selected topic, and message or goals, with your consent. The form handlers send the enquiry through Resend to relevant Promptective mailboxes; they do not write the enquiry content to our application database.

To prevent abuse, we keep short-lived request counters keyed by SHA-256 hashes derived from your email address and source IP address. Each counting window lasts 10 minutes, and stale counters are deleted during later form processing once they are more than 24 hours old.

05

Public website analytics

Public marketing routes use regional analytics controls. For visitors detected in Australia, Google Tag Manager and configured analytics tags are enabled by default after a brief notice, with an immediate option to turn them off. They remain blocked until an affirmative choice in the European Economic Area, United Kingdom, Switzerland, and Crown Dependencies. Missing, malformed, or otherwise unevaluated location information also remains opt-in. Visitors browsing without JavaScript do not receive a GTM fallback. A browser Do Not Track or Global Privacy Control signal keeps analytics off in every region.

The GTM container can run configured analytics tags. Depending on the active tags and events, Google may receive:

  • Your IP address and an approximate location derived from it;
  • The full page URL, including query parameters, page title, and referring URL;
  • Browser, operating system, device, language, and screen information;
  • Visit times and interaction events configured in the container;
  • The analytics storage state applied for your region and preference.

Advertising storage, advertising user data, and advertising personalisation remain denied. Avoid putting personal information in website URLs because configured analytics can receive the full URL while analytics are on.

06

Billing data

Paid subscriptions are processed by Stripe. Payment card details are entered directly with Stripe and never reach our systems. We store Stripe customer and subscription identifiers, package and entitlement state, and licensed endpoint quantities to operate your subscription.

07

Product security telemetry

When an organisation deploys Promptective, our clients and APIs generate routine security records that are deliberately metadata-first:

  • Interaction events: surface, direction, protocol, operation, timestamps, subject and resource identifiers, transport metadata, and content metadata limited to lengths and cryptographic hashes;
  • Findings: detector identifier and version, category, severity, confidence, character-offset spans, and content-free evidence;
  • Policy decisions: outcome, matched rule identifiers, reason codes, applied transformations, and fail mode;
  • Traffic records: request and response timestamps, model identifiers, token counts, latency, and the enforcement action taken. The gateway stores part counts, byte counts, and a SHA-256 content hash; the agent API stores message role, optional name or tool-call identifier, length, and a SHA-256 content hash by default. The legacy development/testing exception described above can store raw agent request messages when explicitly enabled;
  • Endpoint-flow records: source and destination addresses and ports, domain names, TLS metadata, byte counts, minimised application identity (process name, signing identity, publisher, executable hash, and a hashed path), AI-traffic classification, and the enforcement outcome. The current authenticated endpoint-flow API forces metadata mode and accepts no payload or web-page body fields. Full-capture schema scaffolding exists in source but is not an authorised production endpoint-ingestion path;
  • Runtime audit events covering administrative and API activity: actor, action, resource, source IP address, and timestamps.

On supported inspection paths, prompt and response text can be analysed in real time to reach an allow, audit, sanitise, require-approval, or block decision. Short prompt excerpts used by supported on-device explanations stay within the local application, expire automatically, and are not transmitted in routine telemetry.

08

Device and installation data

To enrol and manage installed clients we process:

  • Installation kind (browser, endpoint, desktop, or CLI), platform, client version, and status;
  • Signed endpoint evidence: operating system version, hardware architecture, and device name;
  • Browser family and version, extension identifier and version, and extension artifact hash;
  • Server-side installation credential hashes with short display prefixes, enrolment token hashes, profile or public-key hashes, and last-seen timestamps. The raw installation credential is returned once and stored by the client in browser-profile or endpoint-protected storage so that the client can authenticate; it is not stored raw in the server credential record;
  • Signed, device-reported desktop posture, such as the protection mode the client reports as applied, runtime health, and licence status. This client report describes the device posture. Separate installed-device evidence establishes operating-system enforcement.

09

Website and browser-extension diagnostics, API error monitoring, and service logs

Public website analytics is described in the section above. Separately, where error monitoring is enabled, a browser error that reaches the marketing error hook is reduced to a fixed route category, a generic error label, and bounded code stack locations before it is sent. Form and input values, request and response content, headers, cookies, query strings, and prompt or response text are excluded, and reports are sent without cookies or credentials. Where Sentry is configured for the Control API, standard server-side request error monitoring can process error messages and stack traces, request methods, route and path details (which can include query strings), route metadata, request headers that may include authentication or cookie values, and parsed cookies. The Sentry SDK can filter some network identifiers, but the Control API currently applies no application-owned header allowlist or before-send scrubber. The exact context depends on the error and SDK processing.

Our servers write operational logs to operate and troubleshoot the Services, and the application logger sends them to Better Stack when that integration is configured. The gateway logger rejects fields named for content, messages, responses, credentials, and similar sensitive values; other service code is designed to log operational facts while excluding customer content. Where Promptwatch crawler logging is enabled, the marketing server forwards requests presenting a recognised AI crawler user-agent. These records contain the request time, method, path, response status and content type, crawler IP address, hostname and user-agent string. Query strings and referrers are excluded.

When a user explicitly chooses Send diagnostic report in our browser extension, the extension sends us a one-off report containing its version and current enrolment, protection mode, policy, and upload-queue status. We associate the report with the authenticated organisation and browser installation and record the source network address. Prompts, responses, page addresses, and file contents are excluded. We use the submitted report to investigate support issues and improve service reliability and security.

10

Routine exclusions and explicit exceptions

Current source implements the following boundaries:

  • Routine production telemetry does not store raw prompt or response text. The legacy raw-content switch is an isolated development/testing exception and is prohibited for production and customer data;
  • Server credential records do not store raw passwords, Promptective API keys, installation credentials, or enrolment tokens. Authentication session records contain opaque session-token values, and clients store their raw installation credential locally so they can authenticate;
  • Google Tag Manager and configured analytics tags remain off until allowed in prior opt-in, unknown, and unevaluated regions. In Australia they are on by default until turned off. The integration keeps Google advertising consent denied;
  • Promptective does not sell personal information or share it for behavioural advertising;
  • The current authenticated endpoint-flow API does not accept web page content or network payloads. Legacy full-capture schema and type scaffolding does not grant production capture authority.

11

How we use personal information

We use personal information to:

  • Provide, operate, and secure the Services, including enforcing the protection policies your organisation configures;
  • Authenticate users, installations, and API clients, and prevent abuse through rate limiting and replay protection;
  • Respond to enquiries, investigate diagnostic reports that users choose to submit, provide support, and improve service reliability and security;
  • Manage subscriptions, entitlements, and billing;
  • Send service communications such as email verification, password resets, and organisation invitations;
  • Analyse how AI crawlers access the public website, based on our legitimate interest in understanding and maintaining its availability to those crawlers;
  • Measure public marketing-site use while optional analytics are on;
  • Comply with legal obligations.

Where the GDPR or UK GDPR applies, we rely on performance of a contract, our and our customers’ legitimate interests in securing workforce AI use, our legitimate interest in understanding AI crawler access to the public website, consent for enquiry forms and optional website analytics, and compliance with legal obligations. For visitors detected in Australia, we collect the limited website analytics described above to understand and improve our public website, subject to the notice and opt-out controls described in this policy.

12

Service providers and disclosures

We share personal information only as needed to run the Services, with the following providers:

  • Railway: application hosting, PostgreSQL database, and Redis cache;
  • Resend: account and enquiry email delivery;
  • Stripe: payment processing;
  • Better Stack: server log management, where configured;
  • Sentry: allowlisted marketing-browser error diagnostics and Control API server request and error diagnostics, where enabled;
  • Promptwatch: server-side AI crawler traffic analysis for the public marketing website, where enabled;
  • Google: Tag Manager and configured website analytics while optional analytics are on under the regional setting and visitor preference;
  • The AI provider your organisation configures for a supported gateway route, currently OpenAI or Anthropic: prompts are forwarded to that provider on your organisation’s behalf, and an upstream API key you supply is stored using authenticated encryption.

We may also disclose information where required by law or to protect the rights, safety, or security of Promptective, our customers, or others. We do not sell personal information.

13

Cookies

Authentication session cookies prefixed “promptective” keep account users signed in and protect authenticated requests. On public marketing routes, a first-party “promptective_analytics_consent” cookie records whether analytics were allowed, declined, or enabled under the Australian opt-out default. It lasts for 180 days so the site can apply the setting on later visits. An Australian default setting is not treated as affirmative consent if a later visit is detected in a prior opt-in region.

While analytics are on, tags configured through Google Tag Manager may set first-party analytics cookies such as “_ga” and “_gid”. Turning analytics off keeps GTM off on later page loads and removes recognised Google Analytics cookies available to the site. You can reopen “Privacy choices” at any time. We set no locale cookie, and the integration keeps advertising consent denied.

14

Data retention

Retention differs by record and deployed service:

  • Account and organisation records remain while needed to provide and secure the account and to meet applicable legal or contractual obligations. The current source does not contain an automatic account-closure deletion workflow; verified requests are assessed through the privacy contact process;
  • Authentication sessions have a seven-day validity period and can be ended earlier by sign-out, password reset, or revocation. Verification and password-reset records carry their own expiry; expiry ends their authority but does not by itself prove that a database row has been physically deleted;
  • Enquiry rate-limit counters more than 24 hours old are deleted opportunistically during later form processing. Enquiry content is sent to Resend and recipient mailboxes, whose configured retention is not enforced by the application database;
  • Enterprise administrators can store separate upper-bound policy values for interaction metadata, sanitised diagnostics, and source-fidelity quarantine. The current source does not contain a telemetry deletion worker that enforces those values, so the settings must not be treated as proof that deletion has run;
  • Resend, Stripe and Promptwatch retain the data they process under their own terms and retention settings;
  • The analytics preference expires after 180 days. If analytics tags are configured, Google retains resulting data under the active property settings and its applicable terms.

We respond to verified privacy requests within the period required by applicable law or an applicable customer agreement. We do not make a universal 30-day deletion promise where no such requirement or agreement applies.

15

Your rights

If you are in Australia, the Privacy Act 1988 (Cth) and the Australian Privacy Principles give you rights to access and correct your personal information, and to complain about how it is handled, first to us and then to the Office of the Australian Information Commissioner (OAIC).

If you are in the European Economic Area or the United Kingdom, you may have rights to access, rectify, erase, restrict, or port your personal data, to object to processing, to withdraw consent, and to complain to your supervisory authority.

To exercise any right, contact privacy@promptective.ai or use the privacy contact form. We verify the identity of requesters before acting. Where your data is processed on behalf of your organisation, we may refer the request to your administrator and assist them in responding.

16

Security

We protect personal information with encryption in transit, salted password hashing, hashed server-side storage of Promptective API keys, installation credentials, and enrolment tokens, expiry and access controls for authentication sessions, authenticated encryption for upstream API keys, cryptographically signed policy bundles, replay-protected client requests, organisation-scoped access controls, and audit logging. No method of transmission or storage is completely secure, and we cannot guarantee absolute security.

17

International transfers

Our service providers may process personal information in the United States and other countries. Transfer locations and safeguards depend on the provider, deployment, and customer agreement. Where applicable law requires safeguards, we use the contractual or other transfer measures applicable to that processing; contact us for deployment-specific details.

18

Children

The Services are business software and are not directed to individuals under 16. We do not knowingly collect personal information from children.

19

Changes to this policy

We may update this policy from time to time. We will post the updated policy on this page and revise the “Last updated” date. We will give additional notice where applicable law or a customer agreement requires it.

20

Contact us

For any privacy question, request, or complaint, contact our privacy team at privacy@promptective.ai or use the privacy contact form below.

Privacy contact form

Privacy Policy | Promptective