New report: Early-stage AI governance
➤ Download the report below. And see more resources about this framework here.
➤ In the report:
🎁 A method for evaluating and improving AI responsibility, building on Dotan et al. (2024) and the NIST AI Risk Management Framework.
🎁Detailed examples of governance evaluations from early-stage projects developed at the AI4Gov masters program, as part of a workshop/competition. Thank you to Marzia Mortati and Martina Sciannamè for hosting me!
🎁 The examples include risk triage, evaluating current governance activities, and improvement plans.
🎁Insights arising from the examples, including top prioritized risks, common governance strengths and weaknesses, and strategic plans for early-stage projects
➤ Context:
👉Figuring out how to govern AI responsibly, maximize the benefits, and minimize the risks, is a barrier for many organizations.
👉For example, in a recent BCG survey, 52% of executives said that they actively discourage generative AI adoption. The lack of a responsible AI strategy was the second most common reason.
👉These concerns are appropriate because the stakes for organizations are high, including risks and rewards to the quality of their products, reputation, client attraction, employee attraction, and compliance.
👉 The stakes for end users, related communities, and society at large are also high, including mass disinformation, discrimination, privacy violations, and physical and psychological harm, to name just a few.
👉The report focuses on the earliest stage of the development life-cycle, the ideation phase, demonstrating how AI responsibility is crucial even at this stage.
➤ I'm looking for additional venues for this workshop and pilot partners for the framework. Get in touch if you're interested!
Commenti