Responsible AI Governance - Azure OpenAI GPT: 6-Week Assessment

RSM will perform a risk based customized Responsible AI Governance assessment for the OpenAI GPT-4 model. The engagement aims to understand the organization’s risk appetite for the generative AI technologies, governance structure, and risk associated with each use case. The assessment will include actionable next steps that can effectively help manage the risk associated with such technologies.

The main objective of the AI governance engagement is to produce a risk assessment that can assist your organization in developing and implementing OpenAI GPT-4 in a responsible and reliable way. Our customized Responsible AI Governance assessment includes components of various global guidance and frameworks including Microsoft’s Responsible AI Standards. Our assessment incorporates the following domains:

• Governance, Values and Ethics: Require human oversight and usage authorization on a “need to use” basis only. Incorporate the organization’s values and ethics into the design and implementation of the system. Develop procedures and policies to govern AI.
• Bias: Data leveraged for AI systems are equitable and unbiased.
• Consistent: Train AI systems to produce consistent results.
• Privacy and Data Protection: Incorporate data privacy rules and regulations into the design and implementation of the AI system.
• Security: Protect AI systems from cyber and physical attacks.
• Accountable: Clearly define ownership of the model, its input and outputs. An audit trail of key input & updates should be captured and retained.
• Explainable: The stakeholders should be able to understand the input and explain the model’s output.
• Regulatory Compliance: Factor applicable regulations into the design and implementation of the AI system.

At the conclusion of this assessment, the next step would be to consider a recommendation of the assessment before the OpenAI GPT-4 model is implemented in the production environment.

Project Outline

• Conduct stakeholder Azure OpenAI and GPT-4 AI Governance readiness workshop.
• Gain stakeholder commitment. It is crucial to involve stakeholders in the AI governance risk assessment process.
• We will leverage our customized framework to conduct the assessment. Our frameworks incorporate Microsoft’s Responsible AI Standards. There are a number of different AI governance risk assessment frameworks available. These frameworks can be used to guide organizations in developing their own AI governance risk assessment programs.

• Identify the risks: The first step is to identify the potential risks associated with the AI system. This includes risks such as bias, discrimination, security vulnerabilities, and ethical concerns.
  
• Assess the risks: Once the risks have been identified, they need to be assessed. This involves determining the likelihood and severity of each risk.
  
• Review the assessment: The AI governance risk assessment should be reviewed on a regular basis to ensure that it is still relevant and effective.
  
• The documentation of the assessment: The AI governance risk assessment should be documented. This comprehensive documentation should include the risks that were identified, the assessment of the risks, and the mitigation controls that were implemented.
  

*Pricing will vary based on scope and available Microsoft funding.

Click here to learn more