AI Governance: Responsible and Trustworthy AI

This slide was presented on CEDS UITalks event. I discuss about the concept of AI Governance, Responsible AI, Trustworthy AI, along with the AI risks and sample incidents.

I also share about 12 Principle of Trustworthy AI from "What You Don't Know: AI's Unseen Influence on Your Life and How to Take Back Control" book.

1. AI Governance: Responsible and Trustworthy AI Eryk Budi Pratama, M.Kom , M.M, CIPM, CIPP/E, FIP AI Governance Professional Chairman - Institute of Digital Trust Indonesia (IODTI ) Chairman - Center for Intelligence and Defense Strategies (CIDS) CEDS UI Talks 28 September 2024 [email protected] [email protected] Iodti.org

2. Artificial Intelligence (AI) encompasses many technologies that work together to build innovative solutions that transform society and business. [email protected] Iodti.org Intelligent Automation Predictive Analytics Speech Recognition Virtual Assistants Machine Learning Conversational AI Vision AI AI Simulation Generative AI Deep Learning Autonomous Systems Quantum Artificial General Intelligence (AGI)

3. Global Index – Responsible AI 2024 Report Global Index on Responsible AI adopts a multifaceted approach to measurement in order to generate insights on the performance and competencies of the Responsible AI ecosystem within each country across the 19 thematic areas and 3 dimensions. Indonesia is on rank 67. [email protected] Iodti.org Reference: https://www.global-index.ai /

4. OECD AI Incidents Monitor (AIM) The OECD AI Incidents Monitor (AIM) documents AI incidents and hazards to help policymakers, AI practitioners, and all stakeholders worldwide gain valuable insights into the risks and harms of AI systems. [email protected] Iodti.org Reference: https:// oecd.ai/en/incidents Human Rights, Digital Security, Privacy & Data Governance, and Transparency are the most AI Principle violated in the Top Countries including US and UK. AI Incidents can harm human not only non-physical harm, but also Death, Injury, and Hazard.

5. Sample AI Incidents | AI Deepfake of Brian May Exploited in Scam Offering Fake Queen Backstage Tickets (2024) [email protected] Iodti.org Reference: https://incidentdatabase.ai/apps/incidents / ; https:// www.express.co.uk/entertainment/music/1948801/Brian-May-AI-scam-Queen-backstage-tickets Scammers created an AI-generated deepfake of Queen guitarist Brian May, posting a video on TikTok in which the fake May offers backstage tickets to a Queen concert. The real Brian May warned fans about this "disgusting" scam, emphasizing that Queen has no tour dates planned and does not sell backstage access.

6. Sample AI Incidents | Uber AV Killed Pedestrian in Arizona (2018) [email protected] Iodti.org Reference: https://incidentdatabase.ai/apps/incidents / ; https:// www.nytimes.com/interactive/2018/03/20/us/self-driving-uber-pedestrian-killed.html A woman was struck and killed on Sunday night by an autonomous car operated by Uber in Tempe, Ariz. It was believed to be the first pedestrian death associated with self-driving technology . The vehicle was going about 40 miles an hour on a street with a 45-mile-an-hour speed limit when it struck Ms. Herzberg, 49, who was walking her bicycle across the street, according to the Tempe police.

7. Trustworthy AI | 12 Principles To be considered trustworthy, AI systems should meet these 12 principles and employ a four-step framework to ensure the use of AI is ethical, lawful and robust. [email protected] Iodti.org Reference: https:// www.techtarget.com/searchenterpriseai/tip/What-is-trustworthy-AI-and-why-is-it-important Humane Evaluate  if the use of AI serves humanity or could cause more harm than good to society, the environment and an individual's pursuit of life, freedom and happiness. Private & secure Keep all information used to develop, train, deploy, manage and govern the AI system private and secure. Consensual Seek permission from individuals, business partners and third parties to use their data for AI under development Transparent  Inform any individual who might be affected by the developed AI -- in language understandable to them Accessible Document the decisions the AI made regarding any individuals whose data was used, and make the information available Fair and quality data  Ensure the data used to train and develop the AI system is based on sound data standards and has been thoroughly analyzed and adjusted for biases, bad data and missing data. Accountable Declare , train and communicate the people responsible for fixing the AI system if it malfunction. Traceable Set up monitoring tools, processes and employees to communicate which part of an AI system went wrong and when it happened. Agency-imbuing Set up and communicate an appeals program for any individuals who feel the algorithm's recommendations or source data about them might be incorrect. Feedback-incorporating  Provide ways for users, affected people and experts to offer input into the AI system's ongoing learning. Explainable Explain the AI's decisions and sources in plain wording . Governed and rectifiable Implement model drift and data drift monitoring tools and processes, as well as designated people, to detect if the AI system fails or becomes unsafe, biased or corrupt

8. Trustworthy AI Framework The following framework details the four essential steps organizations should follow to integrate trustworthy AI into their technology initiatives. [email protected] Iodti.org Reference: https:// www.techtarget.com/searchenterpriseai/tip/What-is-trustworthy-AI-and-why-is-it-important

9. AI Governance | Illustrative Operating Model [email protected] Iodti.org Business Units AI Oversight & Governance Chief Information Office Data Science Talent Data Risk / Legal Office Data Science Infrastructure AI Risk Monitoring & Surveillance AI Technology, Standards & Playbook External Data Science Capability (Vendors, Talent, Markets, Ecosystem) AI Use Cases / Teams BU Risk, Legal and Compliance Risk & Controls Traceability AI Project Inventory Shared Platform Data Set Inventory Algorithm Library Risk Management Legal Other Corporate Functions Operational Risk Model Risk Reputation Risk Other Risk Domains Intellectual Property M&A (Data Diligence) Contracts & Agreements Other Legal functions Sales and Marketing HR/Talent Finance Supply Chain/ Procurement Other Corp. functions Business Units drive AI use cases along with localized oversight and risk management processes Businesses leverage other corporate functions, including Legal and Risk, for areas needing specialized support Businesses benefit from the CIO being a common point of reference for all things AI (i.e., data, capability, technology, risk & governance) Data Science Talent / Technology