Legal

Privacy Policy

Effective date: March 22, 2026 · Last updated: November 1, 2026

Version 2026-11-01.1

What changed on November 1, 2026: Section 9 now says how long we keep the record of your acceptance of our terms (for the life of your workspace, with IP addresses and browser information deleted after two years), Section 21 points to the published archive of earlier versions, Section 3 now lists the record we keep when you accept our terms (the time, the document versions, your IP address and browser), and references to the Data Processing Agreement now use its new name, the Data Processing Addendum. What changed on October 31, 2026: Intended is an authority layer only. It decides and signs; it does not execute actions in your systems, run connectors, or store credentials for your third-party systems, and we removed the descriptions of connector execution and connector credentials. What changed on October 29, 2026: we now describe exactly how Intended uses AI (deterministic by default; optional AI features off until you turn them on; no customer data sent to third-party AI providers except through your own provider key), what we store for each decision, how retention and legal holds work, and a new Section 4B on product improvement (which may use machine learning, including models we train and host ourselves, never sent to a third-party AI provider), including a Tier 1 setting for de-identified structural signals that is on by default for Free and Platform workspaces and that you can turn off with one switch. Prompts are never used for product improvement unless you separately turn that on. Intended had no customers when these changes were made, so they take effect on this date. Future material changes follow Section 21.

1. Overview

This Privacy Policy describes how Intended, Inc. ("Intended", "we", "us", or "our") collects, uses, and protects information when you use our website (intended.so), APIs, SDKs, CLI tools, and platform services (collectively, the "Services"). Intended processes minimal data required for AI authority evaluation, audit integrity, and compliance evidence generation. This policy applies to all users of the Services, including visitors to our website, customers, and AI agents authorized by customers.

1A. EU legal representative and GPAI deployer status

Intended is in the process of designating a representative in the European Union pursuant to Article 27 of the GDPR and a representative in the United Kingdom pursuant to the UK GDPR. Until the designated representatives and their contact details are published, EU and UK data subjects may contact eu-privacy@intended.so for all representative-related matters.

2. Lawful bases for processing

In accordance with GDPR Article 6, Intended processes personal data under the following lawful bases:

Contract performance (Article 6(1)(b)): Processing necessary to provide the Services you have subscribed to, including authority evaluation, token issuance, audit trail generation, and account management
Legitimate interests (Article 6(1)(f)): Processing necessary for platform security, fraud prevention, abuse detection, service improvement (including the product-improvement process in Section 4B), and aggregated analytics, where such interests are not overridden by your data protection rights
Consent (Article 6(1)(a)): Where we process data based on your consent (e.g., marketing communications), you may withdraw consent at any time without affecting the lawfulness of prior processing
Legal obligation (Article 6(1)(c)): Processing necessary to comply with applicable laws, regulations, or lawful government requests, including tax, accounting, and regulatory retention requirements

3. Data we collect

Account data

Name and email address for account creation and communication
Company name and role for enterprise account provisioning
API keys (hashed) for authentication and access control
A record of your acceptance of our terms: the time, the document versions, your IP address and browser
Operator Console authentication data processed by Auth0 (operated by Okta, Inc.) at login.intended.so: passkey/WebAuthn credential identifiers (Console login is passkey-primary) and, if you link identities, account-linking metadata for your same-email Google and password identities

Authority decision data

Intent metadata: action type, target system, environment, and risk parameters as classified by the Open Intent
Risk evaluation results: 8-factor scores, risk tier classification, and policy outcomes
Authority tokens: signed JWT claims (action, target, scope, expiry, nonce)
Audit records: hash-chained decision entries with SHA-256 integrity
Escalation records: approver identity, rationale, and timestamp
Runtime integration artifacts: generated policy bundles, inferred presets, endpoint metadata, and related configuration produced when customers use runtime adapter workflows such as the NVIDIA OpenShell integration
Action descriptions and metadata your agents submit. Tool-call argument values are stored only as a cryptographic digest and length, and any argument value that appears in an action description is removed before it is stored. Other text you write in an action description is stored as submitted
For optional plain-English intake: a cryptographic digest of the request text; the structured classification result; and, only if you turn on raw prompt logging, a pattern-redacted copy of the request text encrypted under a key specific to your workspace
Outcome reports: if your agents report what happened after acting on a decision, the outcome status and related metadata you submit
Decision provenance: which classifier, classifier version, knowledge-pack version and rule-set version (with a hash of the exact rules evaluated) made each decision, recorded on every decision path and shown in its receipt

Usage data

API request metadata (endpoint, method, status code, latency)
Decision volume and usage metrics
Error logs for debugging and service reliability

Website analytics

Standard web analytics (page views, referrer, browser type) — no third-party tracking pixels
Form submissions (contact, demo requests) — processed by Intended directly
Cookie data as described in our Cookie Policy

4. Data minimization

Intended adheres to the principle of data minimization. We collect and process only the personal data that is strictly necessary for the purposes described in this policy. Authority decision data is limited to the intent metadata required for policy evaluation and the minimum claims needed for token issuance. Intended does not execute actions in your systems and does not collect or store credentials for your third-party systems; your own agents perform actions, so the results of those actions are not sent to Intended unless you choose to report an outcome. Tool-call argument values are stored only as a digest, and argument values are removed from action descriptions before they are stored. Plain-English request text is stored only as a digest unless you turn on raw prompt logging. Other text in action descriptions is stored as submitted, so we recommend that you keep secrets and personal data out of it. We regularly review our data collection practices to ensure continued compliance with minimization principles.

4A. How we use AI, and data not used for model training

Intended's authority decisions are made by deterministic rules, your policies and your people. No AI model makes, scores or influences a decision unless you turn on one of the optional AI features below. Each is off by default and can be turned on only by an owner or admin of your workspace in Settings → Privacy & AI. A plain-language description is at intended.so/developers/docs/security/ai-use, and the data flow with AI providers is at intended.so/legal/ai-subprocessors-data-flow.

AI-assisted classification (optional, off by default): for plain-English requests only. When it is on, the request text is sent to the provider you choose (OpenAI or Anthropic) using your own API key, and your agreement with that provider also applies. Intended never uses its own API key for this: without your key, requests are classified by the deterministic rules. The model only suggests a classification; your policies still make the decision
AI policy drafting (optional, off by default): your plain-English description is processed by a model Intended hosts in its own infrastructure, and the suggested rules are shown to a person to review. Nothing is saved or enforced unless a person saves it. Intended does not store the description or the draft
Hosted LLM gateway: if you route your own AI traffic through it, it carries that traffic to your chosen provider using your own credentials. Intended does not use AI to decide on that traffic and does not store your prompts or the model's responses
MCP server and MCP gateway: Intended sends nothing to an AI provider. What Intended's tools return (such as decisions, rationale, recorded action text, receipts and policy packs) enters your own AI conversation and is handled under your agreement with your AI provider
Block all AI: one setting in Settings → Privacy & AI turns off every optional AI feature for your workspace. Every change to these settings is recorded in your hash-chained audit log

Intended does not send your data to any third-party AI provider. The only third-party AI processing is when you connect your own provider key for AI-assisted classification, which is your choice and is governed by your agreement with that provider. Intended itself uses your data to improve its products only as described in Section 4B, and anything Intended learns that way reaches the default decision path only as a versioned deterministic rule set, never as an AI model deciding for your agents. Intended will give at least 90 days' advance notice, with a right to object, before any change that sends your data to a new recipient (a new AI provider or any new sub-processor) or changes where your data is processed or transferred. Improvements to models and pipelines that Intended hosts itself, operating within the contribution settings in Section 4B, are not such a change.

4B. Product improvement

To improve classification accuracy and build product intelligence, Intended may use machine learning, including training its own models, on data from workspaces whose contribution setting allows it. This work runs on models Intended hosts in its own infrastructure: Intended does not send your data to any third-party AI provider (such as OpenAI or Anthropic) for product improvement. This is separate from decisions: by default your agents' actions are governed by deterministic rules, and improvements reach the default decision path only as new, versioned deterministic rule sets. Every decision receipt shows which classifier and rule-set version made the decision. We do not sell this data.

Tier 0, your workspace only: your workspace's own records may be used to improve how the Services work for your workspace. This stays within your workspace and is part of providing the Services
Tier 1, contribution to improve Intended for all customers: de-identified structural signals only, with your workspace, people and agents removed: Open Intent codes, decisions and outcomes, argument names (never argument values, which are never stored), timing, and tool and target identifiers. Tier 1 is on by default for Free and Platform workspaces and off by default for Enterprise workspaces. An owner or admin can turn it off (or on) at any time with one switch in Settings → Privacy & AI, and turning on Block all AI also turns it off. Turning it off stops contribution from then on; data already contributed is de-identified and cannot be linked back to your workspace
Prompts: never used for product improvement unless an owner or admin separately turns on "Include redacted prompts in product improvement" in Settings → Privacy & AI. That setting is off by default on every plan and can be turned on only while raw prompt logging is on and Block all AI is off. When it is on, prompts are redacted again before any use (personal data and secrets such as email addresses, phone numbers, URLs, keys and tokens are removed) and are processed only on models Intended hosts itself
Tier 2, anything beyond Tier 1 (for example, using identifiable examples from your records): only under a separate written agreement with you
Ownership: Intended owns the resulting models, rules and de-identified aggregate data. You own your authority decision history

5. How we use data

Provide, maintain, and improve the Authority Runtime and related Services
Evaluate AI intents, calculate risk scores, and enforce policy decisions
Generate and maintain hash-chained audit trails for compliance evidence
Issue and verify cryptographically signed authority tokens
Handle escalation workflows and record human oversight decisions
Communicate with you about your account, support requests, and product updates
Improve classification and build product intelligence as described in Section 4B
Monitor platform health, security, and performance
Comply with legal obligations and respond to lawful requests

6. Third-party runtime integrations

Some Intended features generate artifacts or configuration intended for customer-operated third-party runtimes and systems. For example, Intended may generate policy YAML or scoped integration output for a runtime such as NVIDIA OpenShell / NVIDIA NemoClaw. In those cases, Intended processes the data required to generate the artifact and preserve auditability within Intended. Once that artifact is exported or applied in a customer-operated third-party runtime, the customer's direct relationship with that runtime or provider governs the subsequent processing. Customer-configured third-party runtimes are not Intended sub-processors solely because Intended generated a policy artifact for them.

7. Automated decision-making

The Intended Authority Engine makes automated decisions when evaluating AI agent intents. These decisions include risk scoring, policy matching, and the issuance or denial of authority tokens. Important disclosures about automated decision-making:

Automated decisions are made by the Authority Engine based entirely on policies, risk thresholds, and rules configured by the customer (Controller) — not by Intended
Intended does not unilaterally determine the outcome of authority decisions. The customer defines all policies, risk thresholds, escalation triggers, and approval workflows
Customers may configure escalation workflows that require human review for high-risk decisions, providing meaningful human oversight
Authority decisions do not constitute decisions that produce legal effects or similarly significantly affect natural persons within the meaning of GDPR Article 22, as they govern AI agent actions rather than decisions about individuals
If you believe an automated authority decision has adversely affected you, contact your organization's administrator who controls the policy configuration, or contact us at privacy@intended.so

8. Data protection

Encryption at rest: AES-256-GCM for sensitive data including signing keys. Intended does not store credentials for your third-party systems
Encryption in transit: TLS 1.3 enforced for all API, database, and cache connections
Per-tenant isolation: Every query scoped to tenant ID. No cross-tenant data access surface
Key isolation: Per-tenant signing key pairs (Ed25519 by default; RSA-4096 / ES256 selectable) with encrypted private-key storage
Audit integrity: SHA-256 hash chain makes tampering evident. Any modification breaks the chain and is detectable on verification
Evidence bundles: HMAC-SHA-256 signed packages verifiable via our verification API; portable third-party verification using per-tenant public keys is on the roadmap
Nonce protection: Single-use nonces consumed on first verification. Replay attacks impossible
Argument values: tool-call argument values are stored only as a SHA-256 digest and length
Plain-English requests: request text is stored only as a digest unless you turn on raw prompt logging; if you do, it is pattern-redacted and encrypted with AES-256-GCM under a key specific to your workspace
AI controls: optional AI features are off by default, can be blocked for your whole workspace with one setting, and every change to them is recorded in your hash-chained audit log

9. Data retention

Authority decision data and audit records are retained according to your plan tier. Free plans retain audit data for 30 days. Platform plans retain for 1 year. Enterprise plans support custom retention up to 7 years on encrypted, versioned storage with deletion protection; immutable WORM storage (S3 Object Lock, Compliance mode) for long-term storage is on the enterprise roadmap and available for dedicated deployments. Account data is retained for the duration of your active account plus 30 days after deletion. The record of your acceptance of our terms is kept for the life of your workspace as evidence of our agreement, except that the IP addresses and browser information in it are deleted two years after the acceptance. Usage and analytics data is retained for 90 days for operational purposes.

Retention is enforced by a daily deletion job. Enterprise workspaces without a custom window keep authority decision data for 7 years. If a legal hold is active for your workspace, no data is deleted while the hold is in place. Audit records are deleted from the oldest end of the audit chain so that the remaining chain still verifies.

9A. Cross-context data isolation and CPRA 'sharing' disclosure

Intended operates two separate data contexts: intended.so (marketing website): collects visitor cookies, analytics, contact form data, and apps.intended.so console (product): collects account data, authority decisions, policy data, audit logs. Intended does NOT link, combine, or cross-reference data between these contexts for behavioral advertising, profiling, or targeting purposes. Each domain maintains separate analytics, separate retention schedules, and separate access controls. Under California Consumer Privacy Act (CPRA), "sharing" means selling or sharing personal information for behavioral advertising. Intended does NOT sell or share your information. Cross-context data isolation ensures no behavioral profiling across marketing/product domains.

10. Global privacy control and do not sell or share

Intended respects Global Privacy Control (GPC) signals sent by your browser or device. Users who have enabled GPC will not be subject to: behavioral tracking or cross-site profiling, sale or sharing of personal information for commercial purposes, or targeted advertising based on browsing history. California residents may opt out of the sale or sharing of personal information by selecting "Do Not Sell or Share My Personal Information" in your account privacy settings or by emailing privacy@intended.so with "CPRA Opt-Out" in the subject line. Requests are processed within 45 days and do not incur a service fee.

12. Do Not Sell My Personal Information

Intended does not sell your personal information. We do not sell, rent, or trade personal data to third parties for monetary or other valuable consideration. We do not share personal information for cross-context behavioral advertising. This applies to all users regardless of jurisdiction. If you are a California resident and wish to exercise your rights under the CCPA/CPRA, please see Section 13 below or contact us at privacy@intended.so.

13. Cookies

Intended uses strictly necessary cookies to operate the Services (session management, CSRF protection, and user preferences). We do not use third-party advertising or tracking cookies. For detailed information about the cookies we use, their purposes, and how to manage them, please refer to our Cookie Policy at intended.so/legal/cookies.

14. California privacy rights (CCPA/CPRA)

If you are a California resident, the California Consumer Privacy Act as amended by the California Privacy Rights Act provides you with additional rights regarding your personal information:

Right to know: You have the right to request that we disclose the categories and specific pieces of personal information we have collected about you, the categories of sources, the business purposes for collection, and the categories of third parties with whom we share personal information
Right to delete: You have the right to request deletion of personal information we have collected from you, subject to certain exceptions (e.g., legal retention obligations, ongoing service provision, security purposes)
Right to correct: You have the right to request correction of inaccurate personal information
Right to opt-out of sale/sharing: Intended does not sell personal information or share it for cross-context behavioral advertising. No opt-out is necessary, but you may submit a request for confirmation at privacy@intended.so
Right to limit use of sensitive personal information: To the extent we process sensitive personal information, you may request that we limit its use to purposes necessary to provide the Services
Right to non-discrimination: We will not discriminate against you for exercising any of your CCPA/CPRA rights. We will not deny you services, charge different prices, or provide a different quality of service
Categories of third parties: Intended does not disclose your data to any third-party AI provider. If you connect your own AI provider key for AI-assisted classification, request text goes to that provider at your direction and under your agreement with it (see intended.so/legal/ai-subprocessors-data-flow). Product-improvement data (Section 4B) is de-identified or redacted and is not sold or shared
Retention by category: see Section 9 (authority decision data by plan; account data; usage and analytics data)
To exercise these rights, contact privacy@intended.so. We will verify your identity before processing your request. You may designate an authorized agent to make a request on your behalf

15. Additional U.S. state privacy rights

Residents of certain U.S. states have additional privacy rights under their respective state privacy laws:

Virginia (CDPA): Virginia residents have the right to access, correct, delete, and obtain a copy of their personal data, and to opt out of the processing of personal data for targeted advertising, sale, or profiling. To appeal a denial of a privacy request, contact privacy@intended.so with the subject line "VCDPA Appeal"
Colorado (CPA): Colorado residents have the right to access, correct, delete, and obtain a portable copy of their personal data, and to opt out of targeted advertising, sale, or profiling. Universal opt-out mechanisms are honored
Connecticut (CTDPA): Connecticut residents have the right to access, correct, delete, and obtain a copy of their personal data, and to opt out of the processing of personal data for targeted advertising, sale, or profiling
Intended does not engage in targeted advertising, does not sell personal data, and does not profile consumers for decisions that produce legal or similarly significant effects. These rights are provided as a matter of transparency

16. Your rights (GDPR and general)

Access: Request a copy of your personal data and authority decision history
Correction: Update inaccurate personal data through your account settings or by contacting support
Deletion: Request deletion of your account and associated personal data (subject to legal retention obligations)
Export: Export authority decision data and audit records via the API or evidence bundle export
Objection: Object to specific data processing activities where legitimate interest is the legal basis
Portability: Receive your data in a structured, machine-readable format (JSON or CSV)
Restriction: Request restriction of processing in certain circumstances, such as while a correction request is pending
Withdraw consent: Where processing is based on consent, withdraw at any time without affecting the lawfulness of prior processing

17. Data breach notification

In the event of a personal data breach, Intended will follow these notification procedures:

GDPR: Notify the relevant supervisory authority within 72 hours of becoming aware of a breach that is likely to result in a risk to the rights and freedoms of natural persons. Affected data subjects will be notified without undue delay where the breach is likely to result in a high risk
CCPA/CPRA: Notify affected California residents in accordance with California Civil Code Section 1798.82
Other U.S. states: Comply with applicable state breach notification laws, which generally require notification within 30 to 60 days depending on the jurisdiction
Customer notification: Notify affected customers via email and the platform dashboard within 72 hours of confirming a breach affecting their data
All notifications will include: the nature of the breach, categories of data affected, approximate number of data subjects affected, likely consequences, and measures taken or proposed to address the breach

18. Sub-processors

Intended maintains a current list of sub-processors used to provide the Services at intended.so/legal/subprocessors. The list is also available upon request by contacting dpa@intended.so. We provide advance notice of changes to our sub-processor list in accordance with our Data Processing Addendum. Current active sub-processors include Amazon Web Services (AWS) for infrastructure hosting, Stripe for payment processing, Resend for transactional email, Twilio for SMS delivery (when SMS-based features are enabled), and Auth0 (operated by Okta, Inc.) for Operator Console authentication at login.intended.so (identity, account-linking, and passkey/WebAuthn credential management). Intended uses no third-party AI provider as a sub-processor: AI policy drafting and product improvement run on models Intended hosts in its own infrastructure. If you turn on AI-assisted classification and connect your own OpenAI or Anthropic key, that provider processes your requests under your own agreement with it; this is off by default. See the Subprocessor List for the full, authoritative set. Systems your agents act on and customer-operated third-party runtimes are not Intended sub-processors: Intended does not call or send data to them, and generating an artifact for a customer-operated runtime does not make it a sub-processor.

19. International transfers

Intended processes data in the United States using AWS infrastructure. For customers in the European Economic Area, United Kingdom, or other jurisdictions with data transfer restrictions, we rely on Standard Contractual Clauses (SCCs) and supplementary security measures as described in our Data Processing Addendum. Supplementary Measures for Schrems II Compliance: In addition to Standard Contractual Clauses, Intended implements supplementary technical and organizational measures to address post-Schrems II requirements: (a) Encryption: Personal data is encrypted at rest (AES-256-GCM) and in transit (TLS 1.3). Encryption keys are managed per-tenant and are not accessible to US government authorities without decryption by Intended (key derivation from customer-controlled master key). (b) Isolation: Per-tenant data isolation at application and database levels prevents aggregation of EU data with other datasets, limiting exposure in government access requests. (c) Fail-Closed Architecture: Unresolvable policy evaluations result in authorization denial, not default approval. This prevents over-inclusive government access. (d) Government Access Notification: Intended will notify the Controller of any government access requests for personal data, unless Intended is legally prohibited from doing so. Intended publishes annual transparency reports on government requests. (e) Objection Rights: Upon notification of government access requests, the Controller retains the right to object to processing and may terminate Services if objection is not accommodated. These measures do not eliminate the risk that US authorities may compel Intended to disclose personal data, but they substantially reduce that risk and provide Controller with remedies and visibility. Enterprise customers requiring data residency restrictions should contact us to discuss deployment options.

20. Children's privacy

Intended Services are not directed to individuals under 16 years of age. We do not knowingly collect personal data from children. If we learn that we have collected personal data from a child, we will delete it promptly. If you believe a child has provided us with personal data, please contact privacy@intended.so.

21. Changes to this policy

We may update this Privacy Policy to reflect changes in our practices or legal requirements. We will provide notice of material changes through the Services or by email at least 30 days before the changes take effect. Continued use of the Services after the effective date constitutes acceptance of the updated policy. Every version of this policy that users have accepted is published at intended.so/legal/archive, and earlier versions are available on request.

22. Contact

For privacy-related inquiries, data subject requests, or complaints, contact us at privacy@intended.so or by mail at Intended, Inc., 2261 Market Street, San Francisco, CA 94114, USA. For EU representative inquiries, contact eu-privacy@intended.so. If you are unsatisfied with our response, you have the right to lodge a complaint with your local supervisory authority.