SwarmAI Framework
Copyright 2025-2026 IntelliSwarm.ai

This product includes software developed at IntelliSwarm.ai
(https://www.intelliswarm.ai/).

This product includes software developed by the Spring Framework project
(https://spring.io/) licensed under the Apache License 2.0.

This product includes software developed by the Spring AI project
(https://spring.io/projects/spring-ai) licensed under the Apache License 2.0.

================================================================================
IMPORTANT NOTICES — AUTONOMOUS AI AGENT DISCLAIMER
================================================================================

1. AUTONOMOUS AGENT BEHAVIOR

   SwarmAI is a multi-agent orchestration framework that enables AI agents to
   operate with varying degrees of autonomy, including but not limited to:
   executing tasks, generating and registering new tools (skills) at runtime,
   making decisions based on LLM outputs, and self-improving workflows.

   AI agents powered by this framework may produce unexpected, inaccurate, or
   harmful outputs. The framework provides safety mechanisms (sandboxing,
   approval gates, budget limits, tool permissions) but these do NOT eliminate
   all risk. Users are solely responsible for configuring, supervising, and
   validating agent behavior in their environments.

2. NO WARRANTY FOR AI DECISION-MAKING

   Outputs generated by AI agents — including but not limited to text, code,
   tool invocations, skill generation, and autonomous decisions — are NOT
   warranted for correctness, safety, completeness, or fitness for any
   particular purpose. LLM-based agents are inherently non-deterministic and
   may produce different results on identical inputs.

3. PRODUCTION ENVIRONMENT RESPONSIBILITY

   Deploying SwarmAI in production environments is done entirely at the user's
   own risk. The user is responsible for:
   - Configuring appropriate safety boundaries (tool permissions, budget limits,
     governance gates, sandboxing)
   - Implementing human-in-the-loop oversight appropriate for their use case
   - Testing and validating all agent behaviors before production deployment
   - Monitoring agent actions and outputs in real time
   - Ensuring compliance with applicable laws and regulations

4. SELF-IMPROVING WORKFLOW RISKS

   SwarmAI's self-improving capabilities allow agents to dynamically generate
   code, create new tools, and modify their own behavior during execution.
   While validation and sandboxing mechanisms are provided, dynamically
   generated code may contain bugs, security vulnerabilities, or unintended
   side effects. Users MUST review and approve generated skills before
   promoting them to production use.

5. THIRD-PARTY LLM PROVIDERS

   SwarmAI delegates inference to third-party LLM providers (e.g., OpenAI,
   Anthropic, Ollama). The framework makes no guarantees about the behavior,
   availability, safety, or compliance of third-party model outputs. Users are
   responsible for their own agreements with and usage of LLM providers.

6. DATA HANDLING

   Agents may process, transmit, or store data as part of workflow execution.
   Users are responsible for ensuring that data handling complies with all
   applicable data protection regulations (e.g., GDPR, CCPA) and their own
   organizational policies. The framework does not automatically enforce data
   residency, retention, or classification requirements.

7. SELF-IMPROVEMENT TELEMETRY AND DATA EXPOSURE

   SwarmAI's self-improving module collects execution telemetry — including
   task outcomes, failure patterns, token usage, convergence metrics, tool
   selection data, and skill effectiveness scores — to generate framework
   improvement artifacts. This telemetry collection operates during workflow
   execution and may inadvertently capture, process, or persist data from
   user workflows, including potentially sensitive or regulated data.

   Users are solely responsible for:
   - Ensuring that self-improvement telemetry does not capture data subject
     to regulatory protection (e.g., PII, PHI, financial data)
   - Configuring appropriate data filtering or redaction before enabling
     self-improvement features in environments that process sensitive data
   - Reviewing all generated improvement artifacts (learned policies, skill
     definitions, anti-patterns) for inadvertent data leakage before
     promoting them or contributing them back to the project
   - Ensuring that telemetry data is not transmitted to external systems
     without appropriate authorization and data protection measures

   THE FRAMEWORK DOES NOT AUTOMATICALLY REDACT, CLASSIFY, OR FILTER
   SENSITIVE DATA FROM SELF-IMPROVEMENT TELEMETRY. No representation is
   made that the self-improvement pipeline is compliant with any data
   protection regulation. Users deploying this feature in regulated
   environments do so entirely at their own risk.

8. LIMITATION OF LIABILITY

   TO THE MAXIMUM EXTENT PERMITTED BY APPLICABLE LAW, IN NO EVENT SHALL
   INTELLISWARM.AI OR ANY CONTRIBUTOR BE LIABLE FOR ANY DAMAGES ARISING FROM
   THE USE OF THIS SOFTWARE IN PRODUCTION ENVIRONMENTS, INCLUDING BUT NOT
   LIMITED TO DAMAGES CAUSED BY AUTONOMOUS AGENT ACTIONS, GENERATED CODE,
   SELF-IMPROVING WORKFLOWS, DATA LOSS, SERVICE DISRUPTION, OR ANY OTHER
   CONSEQUENCE OF AI AGENT BEHAVIOR — WHETHER DIRECT, INDIRECT, INCIDENTAL,
   SPECIAL, CONSEQUENTIAL, OR EXEMPLARY.

================================================================================
