📄 KIORI — PRIVACY POLICY

Last Updated: 04.02.2026

Kiori (“we”, “us”, “our”) is a knowledge management and AI assistant platform operated by Crowd Wisdom SL/SLU. We help users retrieve, organize, and augment information using document indexing, search, RAG (Retrieval-Augmented Generation), and agentic AI workflows.

We are committed to protecting your personal data and complying with the General Data Protection Regulation (GDPR), the UK GDPR, DSGVO, and all applicable EU and UK privacy laws.

If you have any questions about this Privacy Policy, you may contact us at: privacy@crowd-wisdom.com

1. Data Controller

Crowd Wisdom SL/SLU (Registered in Spain) Email: privacy@crowd-wisdom.com

For all data processed within the Kiori platform, Crowd Wisdom acts as the Data Controller.

For integrations where users connect their own accounts (e.g., Google Drive, OneDrive), those providers act as independent controllers.

1a. Data Protection Officer (DPO)

Gabriel Michels Crowd Wisdom SL/SLU Email: privacy@crowd-wisdom.com

You may contact the DPO directly for any questions relating to the processing of your personal data or your rights under the GDPR.

2. Types of Personal Data We Process

2.1 Account & Authentication Data

Email address
Password (hashed) if signing up via Email & Password
Google OAuth identifier, email, and optional name/profile picture

Legal Basis: Art. 6(1)(b) GDPR — performance of contract.

2.2 Usage Data & System Logs

We collect limited metadata for security, debugging, and operational purposes:

IP address
User agent
Timestamp & timezone
Session identifiers
Request metadata
Audit logs
Security logs

Legal Basis:

Art. 6(1)(f) GDPR — legitimate interests (security & fraud prevention)
Art. 6(1)(c) GDPR — compliance with legal obligations

2.3 Chat Data, AI Assistant Inputs & Outputs

To enable RAG retrieval, context memory, and agentic workflows, we process:

User chat messages
AI assistant responses
Agentic tool call traces (iterations, actions taken, intermediate queries)
Search queries
Document snippets retrieved via RAG

These are stored so that:

RAG can retrieve past content
You can view your chat history
Models can provide consistent context

Legal Basis: Art. 6(1)(b) — performance of contract.

2.3a User Memory & Chat Memory

To improve the quality and continuity of AI interactions, Kiori extracts and stores contextual memories from your conversations. This enables:

Cross-thread context: The assistant remembers relevant information across different conversations, so you don't have to repeat yourself
Findability: Important decisions, preferences, and facts from past conversations can be recalled when needed
Personalized responses: The assistant adapts to your communication style, domain expertise, and preferences over time
Team alignment: Shared workspace context keeps team members on the same page

Memories are organized at three levels:

Chat memory (thread-scoped): Decisions, constraints, and facts from a specific conversation — enabling continuity within and across threads
User memory (user-scoped): Personal preferences and facts (e.g., preferred language, role, expertise) — enabling personalized assistance across all your conversations
Workspace memory (workspace-scoped): Shared team context (e.g., project conventions, naming standards) — enabling consistent assistance for all workspace members

How it works:

Memories are extracted using AI language models (see §5.1 for providers)
Stored as vector embeddings in our vector database (Qdrant)
Recalled semantically when relevant to your current query

User control:

You can view your stored memories in the application settings
You can delete individual memories or clear all memories at any time
Memory extraction can be disabled per workspace or user preference
Deleting a conversation does not automatically delete extracted memories — these must be managed separately

Temporary agent execution context (short-term memory) is held in memory for up to 5 minutes during active sessions and is automatically discarded.

Legal Basis: Art. 6(1)(b) — performance of contract (enabling personalized AI assistance).

2.4 User-Uploaded Documents & Workspace Data

We store copies of:

Files you upload
Extracted text for embeddings
Embeddings generated for retrieval
File metadata

We also store original documents to allow reprocessing after platform upgrades, which is essential for proper functioning of a knowledge platform.

Legal Basis:

Art. 6(1)(b) GDPR — performance of contract
Art. 6(1)(f) GDPR — legitimate interest in maintaining service integrity and improving retrieval quality

2.5 Potentially Sensitive Data

We do not intentionally process special category data. However, because users may upload arbitrary files, incidental processing may occur.

We employ a PII detection module to flag sensitive elements internally to improve safety and handling. This module may detect:

Personal identifiers
Financial data
Sensitive text snippets

Legal Basis: Art. 6(1)(b) GDPR — performance of contract Art. 6(1)(f) GDPR — legitimate interests (risk reduction & system safety)

We do not profile or analyze users based on sensitive data.

2.6 Payment & Subscription Data

Handled by Stripe:

Email
Plan type
Payment method details
Billing history
VAT-relevant info (if applicable)

Stripe acts as processor/sub-processor.

Legal Basis: Art. 6(1)(b) — performance of contract Art. 6(1)(c) — tax & accounting compliance

2.7 Analytics & Product Insights

We use the following analytics services to understand how users interact with Kiori and to improve the platform:

PostHog

Our primary product analytics tool. PostHog collects:

Page views and navigation events
Signup funnel events (plan selection, account creation, checkout)
Feature usage events (manually tracked, no autocapture)
Cross-subdomain user identification (between www.kiori.co and app.kiori.co)
Exception/error capture

PostHog is configured in identified-only mode (no anonymous person profiles). Data is routed through our proxy endpoint and stored in the EU (eu.posthog.com).

Google Analytics 4 (GA4)

Website analytics including:

Page views
Device and browser data
Session statistics
Signup and conversion events

IP anonymization is enabled.

Vercel Analytics

Performance monitoring on our marketing website, including:

Core Web Vitals
Page load performance metrics

Firebase Analytics

Usage analytics and performance insights for the application.

Analytics are anonymized or pseudonymized whenever possible.

Legal Basis: Art. 6(1)(a) — consent (via cookie banner) for analytics cookies Art. 6(1)(f) — legitimate interest (improving service)

2.8 Advertising (Currently Disabled)

Kiori has integrated Google AdSense to support free public workspaces through advertising. This feature is currently disabled but may be activated in the future.

When enabled, Google AdSense may:

Display banner and interstitial ads to free-tier users on public workspaces
Collect device information, IP address, and browsing context for ad personalization
Set advertising-related cookies (ad_storage)

Paid-tier users are never shown ads.

Legal Basis: Art. 6(1)(a) — consent (ads will only be shown with user consent via cookie/consent mechanisms)

3. How We Use Your Data

We use personal data for the following purposes:

Operating the Kiori platform
Authentication & account management
Enabling RAG retrieval & agentic workflows
Providing AI chat & document search functionality
File storage and reprocessing for service improvements
Payment processing & subscription management
Security, auditing, and fraud prevention
Analytics, usage insights, and performance monitoring
System upgrades that require re-indexing or re-embedding your documents
Customer support and troubleshooting

We do not sell personal data.

4. Legal Bases for Processing

Purpose	Legal Basis
Operating core app features	Art. 6(1)(b)
AI processing & RAG	Art. 6(1)(b)
System logs & security	Art. 6(1)(f), Art. 6(1)(c)
Analytics & cookies	Art. 6(1)(a)
Payments	Art. 6(1)(b), Art. 6(1)(c)
Integrations	Art. 6(1)(b)

5. Data Sharing & Subprocessors

5.1 Subprocessors (LLM & AI Services)

Used for embeddings, generation, reranking, memory extraction, or agentic workflows:

OpenAI (US/EU) — text generation, embeddings
Anthropic (US/EU) — text generation
Google Gemini (EU/Global) — text generation, embeddings
Groq (US) — fast inference, memory extraction
Fireworks AI (US/EU) — reranking, embeddings

These providers operate under GDPR Standard Contractual Clauses (SCCs) when transferring data outside the EEA.

We send the minimal required data for the requested operation (e.g., prompt text, context snippets). LLM providers may change over time as we optimize for quality, speed, and cost. The current list reflects providers actively in use.

5.1a Subprocessors (Web Search & Content Retrieval)

As part of agentic AI workflows, Kiori may access external content on your behalf:

Brave Search (US) — web search queries for agent tool calls
Tavily (US) — AI-optimized web search for agent tool calls
Firecrawl (US) — web page scraping and content extraction
Browserbase (US) — cloud browser sessions for web data collection (e.g., YouTube transcript extraction)

These services receive only the search queries or URLs relevant to the task. No personal user data is shared beyond what is necessary for the request.

5.2 Hosting & Infrastructure

Google Cloud Platform — Netherlands (europe-west-4) Servers, storage, networking, and the following GCP sub-services:
- Cloud KMS — encryption key management for securing user integration tokens
- Document AI — OCR and PDF text extraction
- Google Translate — system prompt translation (not applied to user content)
- Secret Manager — secure storage of platform secrets
Qdrant Cloud — vector storage & similarity search (region dependent)
Firebase (GCP) — authentication, analytics, cloud functions
Vercel — hosting for marketing website, performance analytics
Cloudflare Turnstile — bot protection and CAPTCHA verification (collects IP address, browser fingerprint)

5.3 Payment Provider

Stripe Payments Europe Ltd.

5.4 Email Delivery

Zoho Mail connected to Firebase Auth

5.4a Advertising (Currently Disabled)

Google AdSense (Google Ireland Ltd) — display and interstitial advertising for free-tier users on public workspaces. Currently disabled but infrastructure is integrated.

5.5 Analytics Providers

PostHog (EU) — product analytics, event tracking, cross-subdomain user identification
Vercel Analytics — web performance monitoring
Google Analytics 4 (Google Ireland Ltd) — website analytics
Firebase Analytics (Google Ireland Ltd) — app usage analytics

5.6 Integrations — Independent Controllers

When a user connects external services, these providers become separate Data Controllers:

Google Drive
Microsoft OneDrive / SharePoint (via Microsoft Graph API — accesses files, folders, and metadata with user-authorized scopes)
Notion (via Notion API — accesses pages, databases, and file attachments with user-authorized permissions)

Kiori does not control their data policies; users authorize access directly.

6. International Data Transfers

Because some LLM vendors operate globally, data may be transferred to the United States.

Transfers rely on:

Standard Contractual Clauses (SCCs)
Vendor DPA commitments
Additional safeguards where applicable

We choose EU endpoints when supported but cannot guarantee data always remains in the EU.

7. Cookies

7.1 Cookie Banner

Because analytics may be used (e.g., Google Analytics), we operate a GDPR-compliant cookie consent banner.

7.2 Types of Cookies

Necessary cookies — authentication, session management, bot protection (Cloudflare Turnstile)
Preference cookies — optional
Analytics cookies — PostHog, Google Analytics 4, Firebase Analytics (only with consent)
Advertising cookies — Google AdSense (currently disabled; only with consent when enabled)

See our Cookie Policy for a detailed list of cookies used.

8. Data Retention

8.1 Retention Table

Data Type	Retention	Notes
Account data	Until deletion + 7 days	Grace period for recovery
Uploaded files	Until account/workspace deletion	Auto-cleanup
Audit logs	365 days	Security & compliance
PII detection logs	90 days	Rotated automatically
Security session logs	90 days	Auto-deleted
Subscription/payment data	Legal retention (up to 10 years)	Tax compliance
Chat history	Until user deletes or account deleted	Essential for RAG
User & chat memories	Until user deletes or account deleted	Personalization & context
Short-term agent memory	Up to 5 minutes	Auto-discarded
Agentic traces	Until deletion or account deleted	Supports explainability

Backups may persist up to 30–90 days.

8.2 Reports, Abuse Notices & Legal Requests

Reports, abuse notices, and legal requests submitted through Kiori’s reporting mechanisms may contain personal data such as names, email addresses, organizational affiliation, and supporting documentation. We process this data solely for the purpose of reviewing, responding to, and documenting the reported issue, including compliance with applicable legal obligations. Such data is retained only for as long as necessary to:

investigate and resolve the report,
comply with legal and regulatory obligations,
establish, exercise, or defend legal claims.

Unless a longer retention period is required due to ongoing legal proceedings, report-related data is typically retained for up to 24 months and then securely deleted or anonymized.

9. Your Rights (GDPR)

Users have the right to:

Access personal data
Rectify personal data
Delete personal data (“right to be forgotten”)
Export personal data (portability)
Object to processing
Withdraw consent (for analytics/cookies)
Lodge a complaint with a supervisory authority

We respond to all requests within 30 days.

10. Automated Decision-Making & AI Transparency

Kiori uses:

LLMs for text generation
Embeddings for document search
Agentic workflows for multi-step reasoning

We do not use AI for automated decision-making that produces legal or significant personal effects (Art. 22 GDPR).

Users can always:

delete data
override AI responses
request human assistance

Kiori does not train its models on user data.

11. Data Security

We employ:

Encryption in transit (TLS 1.2+)
Encryption at rest
Strict IAM roles
Audit logs
Secure sandboxing for AI tools
Isolation between tenant workspaces
Regular security reviews

12. Data Deletion

Users can:

Delete chats
Delete documents
Delete their entire account
Export their data before deletion

Once deletion is initiated:

Workspace & documents are removed
Logs tied to identity are anonymized or purged per retention schedule
Backups expire within normal rotation cycles

13. Changes to This Policy

We will update this Privacy Policy as needed. Users will be notified of material changes.

14. Contact

Crowd Wisdom SL/SLU Email: privacy@crowd-wisdom.com