SUPERAGENT USAGE POLICY

Superagent Technologies, Inc. 1111B S Governors Ave STE 3232, Dover, DE 19904

Version 1.0 Effective Date: November 7, 2025 Last Updated: November 7, 2025

Purpose and Scope

This Usage Policy governs acceptable use of Superagent's AI guardrail services (Guard, Verify, and Redact). These guidelines apply to all users accessing the Services through any method including API, SDKs, CLI, Model Context Protocol, or web playground.

This Policy supplements and is incorporated into the Superagent Services Agreement. Capitalized terms not defined here have the meanings given in the Services Agreement.

Our mission is to make AI agents secure and compliant. These guidelines ensure the Services are used responsibly to protect people, respect rights, and advance legitimate security and compliance objectives.

1. Universal Standards

All uses of the Services must comply with these universal standards:

1.1 Legal Compliance

Comply with all applicable laws, regulations, and legal requirements
Do not use the Services for any illegal purpose
Do not process data in violation of export control laws
Do not use the Services in jurisdictions where such use is prohibited

1.2 Safety and Security

Child Safety: Do not use the Services to process, store, or facilitate content involving child sexual abuse or exploitation
Violence: Do not use the Services to facilitate violence, terrorism, or extremism
Weapons: Do not use the Services to develop, produce, or deploy weapons or provide guidance on weapon creation
Malicious Cyber Operations: Do not use the Services for unauthorized hacking, malware distribution, or attacks on systems or networks
Controlled Substances: Do not use the Services to facilitate illegal drug trade or distribution

1.3 Human Rights and Dignity

Harassment: Do not use the Services to facilitate harassment, bullying, threats, or abuse
Hate Speech: Do not use the Services to promote hate speech, discrimination, or violence against individuals or groups
Non-Consensual Content: Do not use the Services to create or distribute non-consensual intimate imagery
Doxing: Do not use the Services to gather or distribute private information for targeting individuals

1.4 Trust and Integrity

Fraud and Deception: Do not use the Services to facilitate fraud, scams, phishing, or financial crimes
Disinformation: Do not use the Services to create or distribute large-scale disinformation campaigns
Spam: Do not use the Services to generate or distribute spam or unsolicited communications
Manipulation: Do not use the Services to manipulate or deceive individuals about the AI nature of outputs
Election Interference: Do not use the Services for voter suppression, election fraud, or interfering with democratic processes

1.5 Privacy and Surveillance

Unauthorized Surveillance: Do not use the Services to track, monitor, or surveil individuals without their knowledge and consent
Biometric Abuse: Do not use the Services to analyze biometric data to infer sensitive characteristics (race, religion, sexual orientation) without proper legal basis and consent
Stalking: Do not use the Services to facilitate stalking or unwanted tracking of individuals

1.6 Intellectual Property

Infringement: Do not use the Services to infringe copyrights, trademarks, trade secrets, or other intellectual property rights
Model Scraping: Do not use the Services' outputs to train competing AI models or services
Unauthorized Use: Do not use the Services to process copyrighted content without proper authorization

2. AI Guardrail Appropriate Use

2.1 Guard (Security Detection)

Designed For:

Detecting prompt injections and jailbreak attempts in AI applications
Blocking unsafe inputs to LLM-powered systems
Identifying malicious tool calls in agentic workflows
Preventing backdoor attacks and security threats
Authorized security testing and red-teaming of AI applications
Protecting AI systems from adversarial inputs

Not For:

General content moderation or censorship of lawful speech
Political content filtering or ideological screening
Surveillance of individuals' communications
Blocking access to information based on viewpoint
Automated decisions about individuals without human review

2.2 Verify (Output Validation)

Designed For:

Validating AI-generated outputs against authoritative sources
Checking factual accuracy of LLM responses
Ensuring policy compliance in AI-generated content
Reducing hallucinations in production AI systems
Quality assurance for customer-facing AI applications
Compliance verification for regulated industries

Not For:

Replacing professional judgment in high-stakes decisions
Automated legal determinations without attorney review
Medical diagnoses without physician oversight
Financial advice without qualified advisor review
Determining eligibility for rights or benefits without human involvement
Making final decisions in regulated domains without human review

2.3 Redact (Sensitive Data Protection)

Designed For:

Identifying and removing PII (personally identifiable information)
Protecting PHI (protected health information) in HIPAA contexts
Detecting secrets, API keys, and credentials in logs
Compliance with data protection regulations (GDPR, CCPA)
Safe logging and monitoring of AI applications
Protecting sensitive information in AI workflows

Not For:

Circumventing data subject rights under privacy laws
Hiding evidence of illegal activity or regulatory violations
Unauthorized collection or processing of personal data
Concealing information required by law enforcement or regulators
Facilitating surveillance without proper legal authority

3. High-Risk Use Cases

If you use the Services for the following high-risk applications, you must implement additional safeguards:

3.1 Required Safeguards

For all high-risk use cases, you must:

Human-in-the-Loop: Maintain qualified human review of all outputs before finalization
Professional Oversight: Ensure licensed or qualified professionals validate outputs in their domain
AI Disclosure: Disclose to end users when AI assistance is used (for consumer-facing applications)
Audit Trails: Maintain logs of AI-assisted decisions and human reviews
Accountability: Establish clear accountability for final decisions

3.2 High-Risk Categories

A. Legal Services

Contract analysis, drafting, or review
Legal research and case law analysis
Compliance advice or guidance
Dispute resolution recommendations

Requirements: Licensed attorney must review all outputs before client delivery. Disclose AI assistance to clients.

B. Healthcare and Medical

Diagnostic assistance or recommendations
Treatment planning or suggestions
Medical record analysis
Health risk assessment

Requirements: Licensed healthcare professional must review all outputs. Comply with HIPAA and medical device regulations where applicable.

C. Financial Services

Investment recommendations
Credit decisions or scoring
Financial planning or advice
Risk assessment

Requirements: Qualified financial advisor must review outputs. Comply with SEC, FINRA, and applicable financial regulations.

D. Employment Decisions

Hiring or firing recommendations
Performance evaluation assistance
Compensation decisions
Promotion or demotion recommendations

Requirements: HR professional must review outputs. Ensure compliance with employment discrimination laws.

E. Housing and Lending

Credit approval or denial
Rental application decisions
Loan underwriting
Risk assessment for housing

Requirements: Qualified underwriter must review decisions. Comply with Fair Housing Act and lending regulations.

F. Education

Student grading or assessment
Admission decisions
Disciplinary recommendations
Academic placement

Requirements: Licensed educator must review outputs. Ensure compliance with FERPA and educational regulations.

G. Criminal Justice and Law Enforcement

Risk assessment for sentencing or parole
Predictive policing applications
Surveillance operations
Investigative analysis

Requirements: Appropriate legal authority and judicial oversight. Comply with constitutional protections and civil rights laws.

4. Professional Services Limitations

You may not use the Services to provide professional services in regulated domains without appropriate licensed professional involvement:

Legal Advice: Requires attorney review and licensing
Medical Diagnosis or Treatment: Requires physician review and licensing
Financial Advice: Requires qualified financial advisor oversight
Mental Health Counseling: Requires licensed therapist involvement
Tax Preparation: Requires qualified tax professional oversight
Engineering Certification: Requires licensed engineer approval

AI-assisted outputs in these domains must be clearly identified as requiring professional review and cannot be presented as final professional opinions.

5. Transparency and Disclosure

5.1 Consumer-Facing Applications

If your application serves consumers (B2C), you must:

Disclose when AI is being used to generate or validate content
Identify that AI outputs may contain errors or limitations
Provide human contact options for disputes or concerns
Maintain appropriate human oversight of AI-generated decisions

5.2 Business Applications

Internal business tools (B2B) are not required to disclose AI use to end employees, but must maintain appropriate governance and accountability structures.

6. Prohibited Activities

You must not:

Technical Misuse:

Reverse engineer, decompile, or disassemble the Services
Attempt to extract model weights or training data
Bypass rate limits, security measures, or access controls
Use automated systems to scrape or mine data from the Services
Benchmark the Services for competitive purposes
Share, sell, or transfer API keys to unauthorized parties
Access the Services to build competing products

Harmful Applications:

Develop or deploy autonomous weapons systems
Create deepfakes for deception or manipulation
Generate spam, phishing content, or malware
Build unauthorized surveillance tools
Target or exploit minors without parental consent
Facilitate election interference or voter suppression
Enable discrimination in housing, employment, or credit

Regulatory Violations:

Violate export control laws or sanctions
Process data in violation of privacy laws (GDPR, CCPA, etc.)
Circumvent industry-specific regulations (HIPAA, GLBA, etc.)
Facilitate money laundering or terrorist financing

7. Special Permissions

7.1 Security Research

Authorized security testing, vulnerability research, and red-teaming of your own AI applications using the Services is permitted with prior written approval from Superagent. Contact security@superagent.sh for authorization.

7.2 Government and Law Enforcement

Government agencies and law enforcement organizations may use the Services for lawful purposes within their jurisdiction and authority. Certain high-risk use cases (surveillance, predictive policing) require additional safeguards and oversight.

8. Regional and Legal Restrictions

The Services comply with U.S. export control laws and may not be available in all jurisdictions
Users are responsible for ensuring their use complies with local laws and regulations
Certain features may be restricted in specific jurisdictions to comply with local law
Users subject to GDPR, CCPA, or other privacy laws must ensure their use of the Services complies with those obligations

9. Monitoring and Enforcement

9.1 Superagent's Rights

Superagent reserves the right to:

Monitor usage patterns for policy violations
Request documentation of safeguards for high-risk use cases
Investigate suspected misuse or violations
Audit customer implementations upon reasonable notice
Review outputs to ensure policy compliance

9.2 Violation Response

If Superagent determines you have violated this Policy:

First Violation (Minor):

Warning and opportunity to cure within specified timeframe
Guidance on compliance measures
Possible temporary feature restrictions

Repeat or Serious Violations:

Immediate suspension of Services
Investigation and cooperation with law enforcement where appropriate
Permanent termination of account and Services
Legal liability for damages caused by misuse

9.3 Appeals

If you believe Superagent has incorrectly determined a policy violation:

Contact support@superagent.sh with explanation and evidence
Superagent will review within 5 business days
Decision will be provided in writing with reasoning
Final appeal may be submitted to legal@superagent.sh

10. Updates to This Policy

Superagent may update this Usage Policy from time to time to:

Address new use cases or capabilities
Respond to regulatory developments
Clarify existing provisions
Strengthen protections for users and third parties

Notice of Changes:

Material changes will be announced 30 days in advance
Notice will be provided via email and dashboard announcement
Continued use after effective date constitutes acceptance
Previous versions available at superagent.sh/usage-policy-archive

Material Changes include new prohibited activities, additional high-risk requirements, or changes to enforcement procedures.

Non-Material Changes (clarifications, examples, organizational improvements) may be effective immediately.

11. Interpretation

If there is any conflict between this Usage Policy and the Services Agreement, the Services Agreement controls except where this Policy provides more specific guidance on acceptable use.

Examples provided in this Policy are illustrative and not exhaustive. The principles and standards apply broadly to similar situations.

12. Contact and Reporting

Questions About This Policy

compliance@superagent.sh

Report Violations

abuse@superagent.sh

Security Issues

security@superagent.sh

General Support

support@superagent.sh

Legal Inquiries

legal@superagent.sh

Superagent Technologies, Inc.
1111B S Governors Ave, Suite 3232
Dover, DE 19904
United States

END OF USAGE POLICY