AI Governance

1. Purpose

This Artificial Intelligence Governance Policy (the "Policy") establishes the principles, controls, and accountability framework governing the design, development, deployment, and operation of artificial intelligence and machine learning systems ("AI Systems") by STUDD AI, LLC ("Company"). This Policy is intended to support lawful, ethical, and responsible AI practices and align with applicable U.S. state artificial intelligence, consumer protection, and data protection laws, including the New York Responsible AI Safety and Education Act (RAISE Act), California automated decision-making and privacy regulations (CPRA/ADMT), and the Colorado Artificial Intelligence Act and similar frameworks ("State AI Laws").

2. Scope

This Policy applies to:

• All AI Systems developed, licensed, deployed, or operated by the Company;
• All employees, contractors, and agents involved in AI-related activities;
• All third-party models or tools integrated into Company products.

3. Guiding Principles

The Company's AI governance program is guided by the following principles:

• Lawfulness: AI Systems will be designed and operated in compliance with applicable laws and regulations.
• Fairness & Non-Discrimination: Reasonable measures will be taken to identify, test, and mitigate unlawful or prohibited bias.
• Transparency: Appropriate disclosures regarding AI use, purpose, and limitations will be provided.
• Human Oversight: Meaningful human review will be available for high-risk AI Systems.
• Accountability: Clear ownership and escalation paths will exist for AI-related decisions.
• Security & Privacy: AI Systems will be developed and operated using appropriate data protection and security safeguards.

4. AI System Inventory & Risk Classification

The Company shall maintain an inventory of all AI Systems, documenting:

• Intended purpose and decision context;
• Data sources and dependencies;
• Whether outputs materially affect legal, economic, or similarly significant rights.

Each AI System shall be classified as low-risk, moderate-risk, or high-risk, with risk classification reviewed upon material modification, retraining, or expansion of use cases.

5. AI Risk & Impact Assessments

Prior to deployment of any moderate-risk or high-risk AI System, the Company shall conduct a documented AI risk and impact assessment addressing:

• Foreseeable discrimination or disparate impact risks;
• Data quality, provenance, and representativeness;
• Privacy, security, and misuse risks;
• Degree of human reliance on AI outputs;
• Availability of human review or override mechanisms.

Risk assessments shall be updated upon material system changes.

6. Data Governance & Model Development Controls

The Company shall implement data governance practices that include:

• Documented data sourcing and provenance;
• Review of training, validation, and testing datasets for representational imbalance or proxy discrimination;
• Data minimization and purpose limitation principles;
• Secure handling of training and inference data.

7. Bias Testing & Mitigation

The Company shall implement documented bias testing procedures proportionate to the risk level of each AI System, which may include:

• Statistical fairness and outcome disparity analysis;
• Error rate comparisons across relevant populations;
• Counterfactual or sensitivity testing;
• Human review of sampled outputs.

Where bias risks are identified, reasonable mitigation measures shall be implemented and documented, including data refinement, feature constraints, retraining, or output calibration.

8. Human Oversight & Escalation

For high-risk AI Systems, the Company shall:

• Define circumstances requiring human-in-the-loop review.
• Provide override or appeal mechanisms where appropriate.
• Train reviewers on intervention criteria.
• Log overrides and corrective actions.

9. Transparency & Customer Communications

The Company shall provide customers with clear, accurate information regarding:

• The use of AI Systems in products or services;
• The intended purpose and appropriate use of AI Systems;
• Known limitations and material risks;
• Availability of human review mechanisms where applicable.

The Company shall avoid representations that AI Systems are error-free or bias-free.

10. Ongoing Monitoring & Model Lifecycle Management

The Company shall monitor AI Systems post-deployment to:

• Detect performance degradation, bias emergence, or model drift;
• Identify feedback loops affecting training data;
• Trigger retraining, modification, or decommissioning where necessary.

Monitoring frequency shall be proportionate to system risk and impact.

11. Third-Party & Customer Controls

Where AI Systems rely on third-party models or are configured by customers, the Company shall:

• Conduct reasonable diligence on third-party AI providers;
• Allocate AI governance responsibilities contractually;
• Prohibit unsupported or unlawful high-risk uses;
• Cooperate with customer and regulatory audits consistent with confidentiality obligations.

12. Training, Accountability & Governance Structure

The Company shall:

• Designate responsible personnel or committees for AI governance oversight;
• Provide periodic training to relevant employees on AI risk and compliance obligations;
• Maintain documentation sufficient to demonstrate compliance with this Policy and applicable State AI Laws.

13. Regulatory Cooperation & Policy Review

The Company shall reasonably cooperate with lawful regulatory inquiries related to AI Systems. This Policy shall be reviewed periodically and updated to reflect changes in law, technology, and industry standards.

14. No Guarantee of Outcomes

AI Systems are probabilistic by nature. This Policy does not require error-free or bias-free outcomes, but rather the implementation of reasonable, good-faith, and risk-based governance measures.

This Policy is intended for internal governance and external diligence purposes and does not create third-party beneficiary rights.

1. Purpose and Scope

This AI Bias Testing & Responsible AI Policy establishes the framework by which STUDD AI, LLC (the "Company") ensures that its artificial intelligence systems, machine learning models, algorithms, and related services (collectively, "AI Systems") are designed, developed, trained, deployed, and maintained in a manner that promotes fairness, transparency, accountability, and non-discrimination.

This Policy applies to:

• All internally developed AI Systems
• All third-party or vendor-provided AI Systems used by or on behalf of the Company
• All AI Systems used in production, testing, or decision-making environments

2. Responsible AI Principles

The Company is committed to commercially reasonable Responsible Artificial Intelligence practices, including:

• Fairness and non-discrimination
• Transparency and explainability appropriate to risk
• Accountability and human oversight
• Risk-based governance and controls
• Compliance with applicable laws and regulations

AI Systems shall not be designed or deployed in a manner that results in unlawful discrimination or unjustified disparate impacts.

3. Bias Testing Program Overview

The Company shall implement and maintain policies, procedures, and technical measures designed to identify, assess, and test AI Systems for bias, unfair discrimination, and disparate impact ("Bias Testing").

Bias Testing is a continuous lifecycle obligation, not a one-time activity.

4. Timing and Frequency of Bias Testing

Bias Testing shall be conducted at the following stages:

• During AI system design, development, and training
• Prior to initial deployment or material use
• On an ongoing basis following deployment
• Following any material modification, retraining, dataset change, update, or change in use case

The scope and frequency of testing shall be proportionate to the risk profile, complexity, and intended use of the AI System.

5. Types of AI Bias Subject to Testing

Bias Testing shall assess, at a minimum, the following categories of bias:

5.1 Data Bias

Bias arising when training or operational data:

• Overrepresents certain demographics
• Underrepresents protected or impacted groups
• Reflects historical inequities or systemic discrimination

5.2 Algorithmic Bias

Bias resulting from:

• Model architecture or design decisions
• Weighting of variables
• Feature selection or optimization methods

5.3 Deployment Bias

Bias occurring when:

• AI Systems are used outside their intended purpose
• Models are applied to materially different populations
• Context or conditions change without appropriate retraining or validation

5.4 Feedback Loop Bias

Bias that compounds over time when biased outputs reinforce biased inputs.

6. AI System Inventory and Risk Classification

6.1 AI System Inventory

The Company shall maintain an inventory identifying:

• All AI and automated decision-making systems
• Applicable use cases and affected populations
• Whether systems are internally developed or vendor-provided

6.2 Risk Classification

Each AI System shall be classified based on risk:

• Low-risk AI - baseline bias testing
• Medium-risk AI - periodic bias testing
• High-risk AI - enhanced, ongoing bias audits

Risk classification shall consider:

• Nature of decisions made
• Impact on individuals or protected classes
• Regulatory exposure

7. Data Review and Validation

Bias Testing shall include evaluation of:

• Data source provenance and legality
• Representativeness of datasets
• Presence of proxy variables correlated with protected traits
• Known data limitations or gaps

Data governance failures shall be treated as a material bias risk.

8. Disparate Impact Analysis

Bias Testing shall assess whether AI outputs:

• Disproportionately affect protected classes
• Produce statistically significant disparities
• Result in unjustified adverse outcomes

Disparate impact analysis shall align with applicable civil rights and anti-discrimination standards.

9. Pre-Deployment Bias Testing

Prior to deployment, the Company shall:

• Test AI outputs across relevant demographic groups
• Simulate edge cases and stress scenarios
• Validate accuracy, fairness, and consistency
• Document results, assumptions, and remediation actions

Deploying AI Systems without documented pre-deployment testing is prohibited absent documented approval.

10. Ongoing Monitoring and Audits

Because AI Systems evolve over time, the Company shall conduct:

• Periodic re-testing
• Performance benchmarking
• Trigger-based reviews following material changes, incidents, or complaints

Monitoring frequency shall scale with risk classification.

11. Bias Mitigation and Remediation

If Bias Testing identifies actual or reasonably foreseeable biased, discriminatory, or materially disparate outcomes, the Company shall promptly implement commercially reasonable Bias Mitigation measures, which may include:

• Model recalibration or adjustment
• Retraining or refinement of training data
• Modification or removal of proxy variables
• Implementation of safeguards or human review mechanisms
• Changes to deployment scope or use

All mitigation actions shall be documented.

12. Vendor and Third-Party AI Obligations

Use of third-party AI Systems does not eliminate responsibility for bias risk.

The Company shall require AI vendors, through contract or policy, to:

• Conduct bias testing
• Maintain reasonable documentation
• Disclose testing methodologies at a high level
• Permit audits or assessments where appropriate
• Cooperate with regulatory inquiries

Vendor compliance shall be monitored based on risk level.

13. Legal and Regulatory Compliance

Bias Testing and Bias Mitigation shall comply with all applicable laws, regulations, and authoritative guidance, including those relating to:

• Artificial intelligence and automated decision-making
• Algorithmic accountability
• Discrimination and civil rights
• Consumer protection

This includes, where applicable:

• New York Responsible AI Safety & Education Act (RAISE Act)
• California Automated Decision-Making Technology (ADMT) regulations
• Colorado Artificial Intelligence Act
• FTC unfair or deceptive practices standards
• Employment and civil rights laws

Enhanced measures shall apply to regulated or high-risk AI Systems.

14. Transparency and Cooperation

Upon reasonable request and subject to confidentiality and trade secret protections, the Company may provide:

• High-level summaries of Responsible AI practices
• Descriptions of Bias Testing and mitigation processes

The Company shall cooperate in good faith to address reasonable concerns relating to bias, fairness, or regulatory risk.

15. Incident Notification

The Company shall promptly escalate and address any material bias-related issue or fairness concern that could reasonably result in legal, regulatory, or reputational risk.

16. Documentation and Recordkeeping

The Company shall maintain documentation sufficient to demonstrate compliance with this Policy, which may include:

• Testing methodologies
• Data sources and assumptions
• Risk classifications
• Testing results and findings
• Mitigation actions taken
• Responsible personnel

Documentation shall be retained in accordance with applicable recordkeeping requirements.

17. Standard of Care; No Absolute Guarantee

The Company recognizes that responsible AI and bias mitigation are evolving technical disciplines and does not guarantee the complete elimination of bias. Nevertheless, the Company shall exercise reasonable care consistent with generally accepted industry standards applicable to AI Systems of similar nature, complexity, and use.

18. State Artificial Intelligence Law Alignment

The Company's AI Bias Testing, mitigation, and governance practices are designed to comply with applicable U.S. state artificial intelligence laws, including the New York Responsible AI Safety & Education Act (RAISE Act), the California Automated Decision-Making Technology (ADMT) regulations, and the Colorado Artificial Intelligence Act, as each may be enacted, amended, or interpreted from time to time.

19. New York Responsible AI Safety & Education Act (RAISE Act) Alignment

For AI Systems subject to or reasonably implicated by the New York RAISE Act, the Company shall:

• Conduct risk-based assessments of AI Systems used in consequential or decision-making contexts
• Implement and document bias testing and impact assessments designed to identify discriminatory or unfair outcomes
• Maintain written governance documentation describing AI system purpose, training approach, testing methodology, and mitigation efforts
• Apply heightened oversight for AI Systems that materially affect individuals' rights, opportunities, or access to services
• Ensure internal accountability and escalation mechanisms for identified bias or safety concerns

The Company shall retain documentation sufficient to demonstrate compliance with the RAISE Act's expectations regarding responsible AI deployment, education, and risk mitigation.

20. California Automated Decision-Making Technology (ADMT) Compliance

For AI Systems that constitute or support Automated Decision-Making Technology under California law, the Company shall implement measures reasonably designed to:

• Identify AI Systems that make or materially assist automated decisions affecting consumers or employees
• Conduct bias and fairness assessments appropriate to the nature and impact of the automated decision
• Support consumer transparency obligations, including disclosures regarding automated decision-making where required
• Enable human review or intervention for decisions that produce legal or similarly significant effects
• Maintain documentation sufficient to support consumer rights requests, regulatory inquiries, and internal audits

Bias Testing conducted under this Policy shall be structured to support compliance with California requirements relating to transparency, fairness, and accountability in automated decision-making systems.

21. Colorado Artificial Intelligence Act Compliance

For AI Systems classified or reasonably anticipated to be classified as High-Risk Artificial Intelligence Systems under the Colorado Artificial Intelligence Act, the Company shall:

• Implement a risk management program designed to identify, assess, and mitigate algorithmic discrimination
• Perform pre-deployment and ongoing bias testing proportionate to the system's risk profile
• Monitor AI Systems for model drift, data changes, and emergent bias
• Maintain records documenting risk assessments, testing results, and mitigation actions
• Apply a duty of reasonable care in the design, development, deployment, and monitoring of AI Systems

The Company shall take commercially reasonable steps to prevent algorithmic discrimination and to respond promptly to identified risks.

22. High-Risk AI Systems - Enhanced Controls

Where AI Systems are used in high-risk or regulated contexts, including but not limited to employment, lending, insurance, healthcare, housing, education, or public benefits, the Company shall apply enhanced controls, which may include:

• Increased testing frequency
• Expanded disparate impact analysis
• Formal approval prior to deployment
• Human-in-the-loop decision review
• Executive or committee-level oversight

23. Regulatory Cooperation and Good-Faith Compliance

The Company shall maintain its AI governance, bias testing, and documentation practices in good faith and in a manner reasonably designed to adapt to evolving regulatory requirements under applicable state AI laws.

Nothing in this Policy shall be interpreted as guaranteeing regulatory outcomes; however, the Company commits to reasonable, risk-based compliance efforts consistent with generally accepted industry standards.

24. Conflict and Interpretation

In the event of a conflict between this Addendum and other sections of the Policy, this Addendum shall govern with respect to AI Systems subject to state artificial intelligence laws. This Addendum shall be interpreted to allow flexibility as laws, regulations, and enforcement guidance evolve.