Terms of Service

Effective Date: February 26, 2026  |  Last Updated: February 26, 2026

1. Definitions

Here's what the key terms in this document mean.

• "Multikor," "we," "us," or "our" — Multikor.ai, Inc., a Delaware corporation, and its affiliates.
• "Services" — All platforms, websites, web apps, mobile apps, APIs, AI chat agents, and related services operated by Multikor, including:
  • Multikor marketing and informational websites
  • Multikor portals and dashboards (each Organization receives a private subdomain)
  • The Multikor AI Chat Agent (web and mobile)
  • The Multikor mobile applications for iOS and Android
  • Administrative and management dashboards
• "User," "you," or "your" — Any individual or entity that accesses or uses the Services.
• "Content" — All text, data, information, files, messages, and materials submitted through the Services.
• "AI Services" — Any artificial intelligence, machine learning, or automated features within the Services, including the AI Chat Agent.
• "Organization" — A company, team, or entity that maintains a multi-user account on the Services.
• "Administrator" — A User with administrative privileges within an Organization, including the ability to invite members, assign roles, and manage feature access.
• "Member" — A User invited to an Organization by an Administrator, whose access is governed by that Administrator.
• "Tenant" — An isolated data environment within the Multikor platform assigned to each Organization. Your data never mixes with another Tenant's data.
• "Small Language Models" or "SLMs" — Multikor's proprietary lightweight AI models (based on architectures like DistilBERT and TinyBERT) that handle classification, routing, and confidence scoring. SLMs decide where your request goes—they don't generate the content you see.
• "Confidence Score" — A numeric value (0–100%) assigned by our SLMs to each decision. Higher scores mean the system is more certain about the right course of action. Confidence scores determine whether a decision is handled automatically or sent to a human for review.
• "Human-in-the-Loop" or "HITL" — The process where a human subject matter expert reviews, corrects, or approves an AI decision. This is the "5%" in our 95/5 model.
• "Self-Healing Workflow" — An automated process that detects errors or low-quality outputs and either regenerates them automatically, queues them for review, or escalates them to a human expert—without requiring you to do anything.
• "Autonomous Action" — Any action the AI takes without real-time human approval, such as classifying a request, routing it to the right model, generating a response, or triggering a self-healing correction.
• "Prohibited Autonomous Decision" — A decision category (medical, legal, financial, employment, or safety-critical) that must always involve human review before being acted upon, regardless of confidence score. See Section 6.8.
• "Data Ingestion" — The asynchronous process of importing data into the Multikor platform from external sources, including CSV/JSON file uploads and API connections to third-party systems.
• "Auto-Schema" — Multikor's automated process that analyzes the full content of imported data to create data schemas, classify fields, detect data types, identify relationships, and flag PII or sensitive information.
• "Third-Party Data Source" — An external system connected to the Multikor platform for data import, such as CRM, ITSM, project management, financial, or ERP systems (e.g., Salesforce, ServiceNow, HubSpot, Jira, Confluence, QuickBooks Online, SAP, Oracle).

2. Acceptance of Terms

By using Multikor, you agree to these Terms.

When you create an account or use any part of the Services—via browser, mobile app, or API—you confirm that you've read and agree to these Terms and our Privacy Policy.

If you're using the Services on behalf of an organization, you represent that you have the authority to bind that organization to these Terms.

We may update these Terms from time to time. We'll give you at least 30 days' notice (via email or in-app) before material changes take effect. If you keep using the Services after that, you're agreeing to the updated Terms.

3. Eligibility

You must be 18 or older to use Multikor.

By using the Services, you confirm that you're at least 18 and have the legal capacity to agree to these Terms.

Some Services (like the Investor Portal and Customer Portal) require prior authorization. Access is restricted to users who've been approved by Multikor.

4. Account Registration and Security

4.1 Account Creation

Keep your account info accurate and your credentials safe.

To use certain features, you'll need to create an account. When you do, you agree to:

• Provide accurate, current, and complete information (name, email, company, role/title)
• Complete email verification when required
• Keep your information up to date
• Keep your login credentials confidential
• Accept responsibility for all activity under your account

4.2 Account Security

You're responsible for keeping your login credentials safe. If you suspect unauthorized access, notify us immediately at security@multikor.ai. We're not liable for losses from unauthorized use of your account.

4.3 Account Management and Deactivation

Accounts are managed centrally by your Organization's Administrator(s).

User access to Multikor is provisioned through administrator invitations. You cannot create an account independently or delete your own account. Account lifecycle—including activation, role assignment, and deactivation—is controlled by your Organization's Administrator(s).

To request deactivation of your account:

• Contact your Administrator: Ask your Organization's Administrator to deactivate your account through the admin dashboard
• Email: If you can't reach your Administrator, contact privacy@multikor.ai and we'll work with your Organization to process your request

When an account is deactivated:

• Your access to the Services is immediately revoked
• Your data is retained per your Organization's retention policies and our data retention schedule (see our Privacy Policy, Section 9)
• Data that is no longer needed is purged according to the applicable retention period
• Data we're legally required to keep (e.g., billing records, audit logs) is retained as required

Deactivated accounts cannot access the Services. Your Organization's Administrator may reactivate your account at their discretion. For questions about data associated with a deactivated account, contact your Administrator or privacy@multikor.ai.

4.4 Organizational Accounts and Administration

Some accounts support multiple users with admin controls.

If your account is part of an Organization, the following applies:

4.4.1 Administrator Responsibilities

If you're designated as an Administrator, you agree to:

• Invite individuals by providing their name, email, company, and job title
• Ensure you have the authority (and, where required, consent) to share the personal information of people you invite
• Assign, modify, or revoke access rights for Members, including access to chat features, monitoring dashboards, conversation logs, and reporting tools
• Take responsibility for Member activity within your Organization and ensure they comply with these Terms
• Promptly revoke access for anyone who leaves your organization or no longer needs it

4.4.2 Member Acknowledgment

If you're a Member of an Organization, you acknowledge that:

• Your Administrator(s) may see your usage data, including chatbot conversations, session activity, and engagement metrics
• Your Administrator(s) may view, export, or analyze your chatbot interactions for quality assurance, compliance, and reporting
• Your Administrator(s) control your access level and may modify or revoke your permissions at any time
• If your access is revoked, your right to use the Services under that Organization ends immediately
• Account deactivation is managed by your Administrator(s); you cannot deactivate your own account
• Data associated with deactivated accounts is retained per your Organization's retention policies and our data retention schedule

4.4.3 Monitoring and Reporting Access

Administrators with the right permissions may access:

• Chatbot Monitoring: Real-time and historical AI Chat Agent conversations for Members
• Usage Analytics: Dashboards showing user activity, session data, and engagement metrics
• Reporting and Export: Reports on chatbot usage, conversation logs, token consumption, and Member activity
• User Management: Add, edit, or remove Members; assign or modify roles and permissions

Administrators must use monitoring and reporting tools only for legitimate business purposes and in compliance with applicable privacy laws. Misusing admin access—including unauthorized surveillance or sharing monitored data outside the Organization—violates these Terms.

5. Permitted Use

Use Multikor for its intended business purposes. Don't misuse it.

You're granted a limited, non-exclusive, non-transferable, revocable license to access and use the Services for their intended business purposes, subject to these Terms.

5.1 You agree NOT to:

• Use the Services for any unlawful, fraudulent, or malicious purpose
• Reverse-engineer, decompile, or disassemble any part of the Services
• Circumvent, disable, or interfere with security features or access controls
• Use automated tools (bots, scrapers, crawlers) without our prior written consent
• Transmit malware, viruses, or other harmful code
• Impersonate another person or entity
• Use the AI Services to generate illegal, harmful, threatening, abusive, harassing, defamatory, or discriminatory content
• Attempt to extract, replicate, or reverse-engineer the AI models, training data, system prompts, or proprietary algorithms
• Use the AI Services to develop competing AI products or services
• Submit sensitive personal data (Social Security numbers, financial accounts, health records) through the AI Chat Agent unless explicitly supported
• Exceed rate limits, usage quotas, or access restrictions on your account
• Share, resell, or redistribute access to the Services or your credentials
• Rely on AI outputs as the sole basis for high-stakes decisions (medical, legal, financial, employment, or safety-critical) without independent human review. See Section 6.8 for details.

6. AI-Powered Services

This section explains how our AI works, what it can do on its own, and where humans stay in the loop.

6.1 How Our AI Architecture Works

The Multikor platform uses a multi-tier AI architecture. Here's how it breaks down:

• Multikor SLMs (classification & routing): Our proprietary small language models, built on architectures like DistilBERT and TinyBERT, handle classification, intent detection, and confidence scoring. They decide where your request goes—they don't write the response you see.
• Third-party LLMs (content generation): Once our SLMs route your request, third-party large language models from leading AI providers (including Anthropic, OpenAI, Google, and others) generate the actual response.
• Confidence-based routing: Every request gets a confidence score. That score determines which tier of LLM handles it and whether a human needs to review it.

You acknowledge and agree that:

• AI-generated content is produced algorithmically and may not always be accurate, complete, or current
• AI responses are not professional, legal, financial, medical, or other expert advice
• You're responsible for evaluating and acting on AI-generated content
• We don't guarantee the accuracy or suitability of AI outputs for any specific purpose

6.2 Conversation Data and How We Learn

We store your conversations and use interaction patterns to make the platform smarter over time.

When you use the AI Chat Agent, your messages and the AI's responses are stored for:

• Maintaining conversation continuity within and across sessions
• Improving the quality and relevance of AI responses
• Refining Multikor's proprietary SLMs through feedback loop processing (see Section 6.3.3)
• Platform analytics and usage monitoring
• Compliance and audit purposes

For full details on data collection, storage, and processing, see our Privacy Policy.

6.3 Human Oversight (The 95/5 Model)

About 95% of decisions are handled by AI; about 5% are reviewed by human experts. Here's exactly how that works.

6.3.1 Three-Tier Decision Routing

Every decision the AI makes gets a confidence score. That score determines what happens next:

6.3.2 Self-Healing Workflows

When the platform detects an error or low-quality output, it self-heals using a three-tier system:

• Tier 1 — Auto-regenerate (~85% of issues): The system detects the problem and regenerates the output automatically. You get a better result without any delay.
• Tier 2 — Quarantine + async review (~10%): The output is flagged and set aside for a human expert to review asynchronously. You may receive a provisional response with a note that it's being verified.
• Tier 3 — HITL escalation (~5%): A human expert steps in to review and correct the issue directly. This is the same pool of experts that handles the "5%" in our 95/5 model.

We also use Gelfand semantic validation to catch hallucinations—a process that checks whether AI outputs are internally consistent and factually grounded before they reach you.

6.3.3 Continuous Improvement: Feedback Loops

This is how the platform gets smarter over time. We run 24 feedback loops across 7 categories:

• When a human expert corrects an AI decision, that correction refines confidence thresholds and routing accuracy
• The SLMs learn from the pattern of the correction (e.g., "requests like X should route to tier Y"), not from your raw data
• Cross-client learning uses only anonymized, aggregated interaction patterns—raw data never crosses Tenant boundaries
• This feedback loop process is what enables the self-healing workflows described above

6.4 Third-Party AI Providers

We use multiple AI providers, selected based on your request's complexity.

Our SLMs route your request to the most appropriate third-party model based on confidence scoring and complexity. Providers currently include Anthropic, OpenAI, Google, and others, with model tiers selected dynamically (e.g., faster/lighter models for routine requests, more capable models for complex ones).

Your use of AI features is also subject to those providers' terms. We're not responsible for the performance, availability, or outputs of third-party models.

6.5 Voice Features

Some AI features support voice input and text-to-speech. When you use voice features:

• Audio may be processed using device-native APIs (e.g., Web Speech API) or third-party speech services
• Voice data may be sent to third-party providers for processing
• You consent to microphone access when prompted by your device or browser

6.6 AI Disclosure Requirements

If the AI acts on your behalf, you need to let people know.

When the AI Chat Agent takes actions on your behalf that affect third parties (e.g., sending communications, submitting forms, or making requests to external services), you must:

• Disclose that AI was involved in generating or sending the communication
• Not represent AI-generated output as being entirely your own work where disclosure is required by law, regulation, or professional standards
• Comply with any applicable AI disclosure laws in your jurisdiction

6.7 Autonomous Action Guardrails

Here's what the AI can and can't do on its own.

The AI CAN do these things autonomously (at ≥ 90% confidence):

• Classify and route requests to the right model or human
• Generate responses to standard queries
• Trigger self-healing workflows to fix errors
• Apply content safety filters
• Log decisions for audit purposes

The AI CANNOT do these things without human review:

• Act on decisions below the confidence threshold (see Section 6.3.1)
• Override a human expert's prior correction
• Access data outside your Tenant boundaries
• Make Prohibited Autonomous Decisions (see Section 6.8)
• Modify its own confidence thresholds or routing rules

Your Organization's Administrator can further restrict what the AI is allowed to do autonomously.

6.8 Prohibited High-Stakes Autonomous Decisions

• Medical: Diagnoses, treatment recommendations, medication decisions
• Legal: Legal advice, contract interpretation, regulatory compliance determinations
• Financial: Investment advice, credit decisions, financial planning recommendations
• Employment: Hiring, firing, promotion, or disciplinary decisions
• Safety-critical: Decisions that could affect physical safety or critical infrastructure
• Regulated industry: Any decision in a regulated industry where human oversight is required by law

Per-tenant Bedrock Guardrails add an extra layer of protection calibrated to your industry (healthcare = maximum, finance = high, retail = standard), but they don't replace the requirement for human oversight on prohibited decisions.

6.9 Content Safety and Quality Assurance

We run multiple safety checks on every AI output before it reaches you.

• Per-tenant Bedrock Guardrails: Content filtering calibrated to your industry and risk profile. Healthcare tenants get maximum filtering; financial tenants get high; retail gets standard.
• Gelfand semantic validation: Checks AI outputs for internal consistency and factual grounding to catch hallucinations.
• Five-category output testing: Every LLM output is evaluated across five quality dimensions before delivery: accuracy, relevance, safety, coherence, and completeness.

6.10 Data Ingestion and Integration Services

You can import data from files or connect third-party systems. Here's how that works and what you're responsible for.

The Multikor platform supports asynchronous data ingestion from multiple sources:

• File uploads: CSV and JSON files uploaded directly through the platform
• API integrations: Connections to third-party data sources including but not limited to CRM systems (Salesforce, HubSpot), ITSM platforms (ServiceNow), project management tools (Jira, Confluence), financial systems (Intuit QuickBooks Online), and ERP platforms (SAP, Oracle)

All data imports are processed asynchronously—you submit the import and we process it in the background, notifying you when it's complete.

Who controls integrations:

• Customers can self-configure certain third-party integrations directly through the platform's admin dashboard
• Some integrations require Multikor assistance for initial setup, credential configuration, or custom field mapping
• Administrators control which integrations are active for their Organization and which Members can trigger imports

Your responsibilities when importing data:

• You are responsible for ensuring you have the right to import the data (including any third-party data, personal data, or regulated data)
• You must comply with the terms of service of any Third-Party Data Source you connect
• You are responsible for the accuracy, legality, and quality of data you import
• You must not import data that you know to be inaccurate, stolen, or obtained in violation of any law or third-party agreement
• If you import personal data of others, you represent that you have the necessary consent or legal basis to do so

Imported data is treated as Customer Data under Section 8.2—you own it, and the same ownership boundaries apply.

6.11 Auto-Schema Creation

When you import data, the platform automatically analyzes its content to build a schema. Here's what that means.

When data is ingested, Multikor's Auto-Schema system performs a full content analysis of the imported data to:

• Detect data types (text, numeric, date, boolean, etc.) for each field
• Classify fields by category (e.g., contact info, financial data, operational metrics)
• Identify relationships between data fields and tables
• Detect and flag PII and sensitive data (names, emails, phone numbers, financial identifiers, health information)
• Generate a structured schema that the AI Chat Agent and analytics tools can query

You should review auto-generated schemas for accuracy, especially for sensitive or regulated data. You can modify, correct, or override schema classifications through the admin dashboard.

7. Mobile Applications

7.1 App Store Terms

If you download Multikor mobile apps from the Apple App Store or Google Play Store, you also agree to those platforms' terms. If there's a conflict between these Terms and an app store's terms, the more restrictive term applies.

7.2 Device Permissions

The mobile apps may request access to:

• Microphone: For voice input to the AI Chat Agent
• Notifications: For service updates and events requiring your attention
• Internet Access: Required for all functionality
• Storage: For caching and offline capabilities

You can manage these permissions through your device settings at any time. Disabling some permissions may limit functionality.

7.3 Updates

We may release app updates from time to time. Some may be applied automatically; others may need your action. Using outdated versions may result in reduced functionality or security issues.

7.4 Apple-Specific Terms

If you access the Services through an Apple device:

• These Terms are between you and Multikor, not Apple Inc.
• Apple has no obligation to provide maintenance or support for the app
• If the app fails to meet any applicable warranty, you may notify Apple for a refund of the purchase price (if any); Apple has no other warranty obligation
• Apple isn't responsible for claims related to the app
• Apple is a third-party beneficiary of these Terms and may enforce them

7.5 Google Play-Specific Terms

If you access the Services through an Android device:

• These Terms are between you and Multikor, not Google LLC
• Google has no obligation or liability regarding the app or these Terms
• Google is a third-party beneficiary of these Terms

8. Intellectual Property and Data Ownership

The short version: you own your data, we own the platform. Here are the details.

8.1 Multikor Platform Intellectual Property

Everything that makes the platform work belongs to Multikor.ai, Inc. (or our licensors). This includes:

• The Multikor name, logos, trademarks, and brand assets
• Proprietary small language models (SLMs), including DistilBERT-based and TinyBERT-based classification and routing models
• The 95/5 model, confidence-based routing engine, and all scoring algorithms
• Self-healing workflow algorithms, agentic automation frameworks, and feedback loop systems
• Model weights, parameters, architectures, and training methodologies
• Per-tenant Bedrock Guardrail configurations and content safety systems
• Analytics, insights, and derived intelligence from aggregated platform usage
• All improvements, enhancements, and derivative works to the platform, regardless of the data that informed them

Nothing in these Terms gives you any ownership interest in the platform, models, or technology.

8.2 Your Data — Customer Ownership

By using the Services, you grant Multikor a worldwide, non-exclusive, royalty-free, sublicensable license to use, process, store, and analyze your Customer Data for:

• Operating and delivering the Services to you
• Refining confidence thresholds and routing accuracy through aggregated, de-identified feedback loop processing
• Powering self-healing workflows through human-in-the-loop expert feedback
• Generating aggregated, de-identified analytics and insights
• Maintaining and improving platform quality, accuracy, and reliability

This license survives termination of your account only for data that has been incorporated into models in aggregated or de-identified form (i.e., data that can't be linked back to you or your Organization). When your account is deactivated, your identifiable Customer Data is handled per your Organization's retention policies and our Privacy Policy.

8.3 The Boundary — What's Yours vs. What's Ours

8.4 AI-Generated Output

Subject to third-party AI provider terms, you may use AI-generated outputs for your lawful business purposes. We don't claim ownership of outputs provided to you, but we make no warranty about their originality or IP status.

9. Confidentiality

Some Services give you access to confidential materials. Please keep them confidential.

Certain Services (including the Investor Portal) may provide access to confidential business information, financial data, and proprietary materials. You agree to:

• Treat all such materials as confidential
• Not disclose or share confidential materials with unauthorized parties
• Use confidential materials only for their intended purpose
• Comply with any additional NDAs you've signed with Multikor

10. Payment and Subscriptions

Some features may require payment. Here's how that works.

If applicable:

• Pricing and billing terms will be presented before purchase
• Subscriptions auto-renew unless canceled before the renewal date
• Refund policies will be disclosed at the point of purchase
• For mobile app purchases, the app store's payment and refund policies also apply

11. Termination

11.1 By You

You can stop using the Services at any time. To deactivate your account, contact your Organization's Administrator (see Section 4.3).

11.2 By Multikor

We may suspend or terminate your access, with or without notice, if:

• You violate these Terms or any applicable law
• Your conduct poses a security risk to the Services or other users
• We're required to do so by law or legal process
• We discontinue or materially modify the Services

11.3 Effect of Termination

When your access ends, your right to use the Services stops immediately. Sections that should naturally survive (Intellectual Property, Limitation of Liability, Indemnification, Dispute Resolution) continue to apply.

12. Disclaimers

Specifically, we don't warrant that:

• The Services will be uninterrupted, error-free, or secure
• AI-generated content will be accurate, complete, reliable, or fit for any purpose
• Defects will be corrected in a timely manner
• The Services will meet your specific requirements

You use the Services at your own risk.

13. Limitation of Liability

Some jurisdictions don't allow exclusion or limitation of certain damages. In those cases, our liability is limited to the fullest extent permitted by law.

14. Indemnification

14.1 Your Indemnification Obligations

You agree to indemnify, defend, and hold harmless Multikor and its officers, directors, employees, agents, and licensors from claims, liabilities, damages, losses, and expenses (including reasonable legal fees) arising from:

• Your use of the Services
• Your violation of these Terms
• Your violation of any third-party rights
• Content you submit through the Services

14.2 Indemnification Carve-Outs

Your indemnification obligation does NOT apply to the extent that a claim arises from:

• (a) Your misconfiguration: Claims resulting from your Organization's custom guardrail settings, routing overrides, or administrative configurations that deviate from Multikor's recommended defaults
• (b) Faulty customer data: Claims caused by inaccurate, incomplete, misleading, or harmful data that you submitted to the Services
• (c) Human-approved escalations: Claims arising from actions taken after a human expert within your Organization reviewed and approved an AI-escalated decision
• (d) Reliance on prohibited autonomous decisions: Claims arising from your reliance on AI outputs for high-stakes decisions (see Section 6.8) without the required independent human review

For clarity: carve-outs (a) through (c) reduce Multikor's indemnification obligation to you. Carve-out (d) reduces your ability to seek indemnification from Multikor when you bypassed required human oversight.

15. Dispute Resolution

15.1 Governing Law

These Terms are governed by the laws of the State of Delaware, United States, without regard to conflict of laws provisions.

15.2 Arbitration

Disputes arising from these Terms or the Services will be resolved through binding arbitration under the American Arbitration Association (AAA) Commercial Arbitration Rules. Arbitration will be in English, in Delaware or remotely, at Multikor's discretion.

15.3 Class Action Waiver

You agree to resolve disputes individually. You waive any right to participate in a class action, class arbitration, or representative proceeding.

15.4 Exceptions

Nothing here prevents either party from seeking injunctive or equitable relief in court for IP infringement or unauthorized access claims.

16. International Users

The Services are operated from the United States. If you're accessing them from outside the US, you're responsible for compliance with your local laws. If any provision conflicts with mandatory local law, it will be modified to the minimum extent necessary.

17. Severability

If any provision is found unenforceable or invalid, it will be limited or removed to the minimum extent necessary. The rest of these Terms remain in full effect.

18. Entire Agreement

These Terms, together with the Privacy Policy and any other agreements between you and Multikor, constitute the complete agreement between us regarding the Services.

19. Contact Us

Questions? We're happy to help.

• Support: support@multikor.ai
• Privacy: privacy@multikor.ai
• Security: security@multikor.ai
• Website: https://multikor.ai  |  https://multikor.com