We provide security updates for the following versions of AgentTune:

Please do not report security vulnerabilities through public GitHub issues.

If you discover a security vulnerability in AgentTune, please report it to us privately by emailing:

security@lexsi.ai

When reporting a vulnerability, please include:

1. Description of the vulnerability
2. Steps to reproduce the issue
3. Potential impact of the vulnerability
4. Affected versions (if known)
5. Suggested fix (if you have one)

• Initial Response: Within 48 hours
• Status Update: Within 7 days
• Fix Timeline: Depends on severity
  • Critical: 1-7 days
  • High: 7-30 days
  • Medium: 30-90 days
  • Low: Next planned release

1. Acknowledgment: We'll confirm receipt of your report
2. Investigation: We'll investigate and assess the severity
3. Updates: We'll keep you informed of our progress
4. Resolution: We'll notify you when the issue is fixed
5. Credit: With your permission, we'll acknowledge your contribution

When using AgentTune, follow these security best practices:

1. Keep Updated: Always use the latest version
2. Validate Inputs: Sanitize user inputs before passing to models
3. Secure Credentials: Never hardcode API keys or tokens
4. Review Dependencies: Regularly audit dependencies for vulnerabilities
5. Use HTTPS: Always use secure connections for API calls
6. Limit Permissions: Run with minimum necessary privileges
7. Monitor Logs: Watch for suspicious activity

When contributing to AgentTune:

1. Code Review: All code must be reviewed before merging
2. Dependency Updates: Keep dependencies up to date
3. Input Validation: Always validate and sanitize inputs
4. Secrets Management: Never commit secrets or credentials
5. Security Testing: Test for common vulnerabilities
6. Documentation: Document security-relevant changes

• Only load models from trusted sources
• Verify model checksums when available
• Be cautious with user-provided model paths

• Sanitize dataset inputs to prevent injection attacks
• Validate file paths to prevent directory traversal
• Limit file sizes to prevent resource exhaustion

• Use API rate limiting where applicable
• Implement proper authentication for services
• Validate all external API responses

• Run training in isolated environments when possible
• Monitor resource usage to detect abuse
• Implement proper access controls

AgentTune depends on several third-party libraries. We:

• Regularly update dependencies
• Monitor security advisories
• Use tools like pip-audit to check for vulnerabilities

To check your installation:

pip install pip-audit
pip-audit

When a security issue is fixed:

1. We will prepare a security advisory
2. We will release a patch version
3. We will publicly disclose the issue after users have had time to update
4. We will credit the reporter (with permission)

Security updates will be:

• Released as patch versions (e.g., 1.0.1)
• Posted as GitHub Security Advisories
• Announced in the CHANGELOG

For security-related questions or concerns:

• Email: security@lexsi.ai
• General Support: support@lexsi.ai
• Website: https://lexsi.ai

We currently do not have a formal bug bounty program, but we:

• Appreciate and acknowledge security researchers
• Provide credit in release notes (with permission)
• May consider rewards on a case-by-case basis for critical vulnerabilities

AgentTune is designed to be used in compliance with:

• Data protection regulations (GDPR, CCPA, etc.)
• Industry security standards
• Research ethics guidelines

Users are responsible for ensuring their use of AgentTune complies with applicable laws and regulations.

• OWASP Top 10
• CWE/SANS Top 25
• Python Security Best Practices

Thank you for helping keep AgentTune and its users safe!