AI Governance

   Principles
      P1 = Safe        = No Weaponization
      P2 = Ethical     = Beyond Legal
      P3 = Accountabie = Responsibile, Liabie
      P4 = Explainable = Transparent
      P5 = Private     = Secure

   Systems
      How Developed
         Stages 
            1. Application                           <-- Governamce
               1.1. Usage: Original vs Actual        <-- Governance
                    1.1.1 Impact                     <-- Governance: P1 = Safe?
                                                                     P2 = Ethical?
                                                                     P3 = Accountable?
                                                                     P4 = Explainable?
                                                                     P5 = Private?
            2. RAG = Retrieval Augmented Generation  <-- Governance: P5 = Private?
            3. Model                                 <-- Govenrance
               3.1. GIGO = Garbage-In-Garbage-Out             ==> Data Quality
               3.2. BIBO = Bias-In-Bias-Out                   ==> Data Diversity, Fair Sampling: Gender, Race, Age
               3.3. FIHO = Facts-In-Hallucination-Out         ==> Data Fact Check: Human-in-the-Loop: Health, Legal, Financial
               3.4. PII = Personally Identifiable Information ==> Data Filter: Remove PII, Redact
                    Copyright data                            
               3.5. Provenance = Source                       ==> Data Labeling Tracking
            4. Prompt                                <-- Governance
               4.1. Intention: Orginal vs Actual
               4.2  Clarity, Specificity
               4.3. Audience
      How Actioned
         If something wrong, what mitigation 
         Due Dilligence
         Human in the Loop


              

|----------------------------------------------------------------------------------------------------------|
| Ethics                          | Data                        | AI Model                                 |
|----------------------------------------------------------------------------------------------------------|
| 1. Interconnected, Distinct     | Collection, Use, Protection | Moral implications of AI-decision making | 
| 2. Accountability, Transparency | Need clear guidelines       | Need clear guidelines                    |
| 3. Bias                         | Unbiased Sampling           | Check model response                     |  
| 4. Privacy, Security            | Regulations, Protection     | Regulations, Protection                  | 
| 5. Governance, Oversight        | Framework                   | Framework                                | 
|   +-- Lazy?                     |                             |                                          | 
|   +-- Incentivization?          |                             |                                          | 
|----------------------------------------------------------------------------------------------------------|