The Road to Accountable AI

In the final episode of our series on governing AI agents, Kevin Werbach speaks with Harish Peri, SVP and General Manager for AI Security at Okta. Peri frames agent governance as the natural next chapter of what Okta has done for two decades: standing in the middle of people accessing technology. The twist is that the new "software" is a non-deterministic agent with a brain, which imposes a much higher security bar. He argues that agents live at the application layer, where the real question is one of authorization: is this agent allowed to take this action or access this data, at this moment, on behalf of this user, given all available signals? Much of the conversation explores why a neutral, independent control plane separate from the frontier models and agent runtimes matters from a cybersecurity standpoint, spreading risk across multiple layers rather than concentrating it in one place. Peri notes that while awareness of rogue AI is universal, roughly 20% of agents carry about 80% of the risk. He distinguishes security threats like prompt injection and poisoned skill files from "intent mismatch," where an under-specified instruction such as "clean this up" gets read as "delete," and explains how coarse-grained limits, fine-grained context-based authorization, and selectively applied human-in-the-loop checks each play a role in agent governance. Harish Peri is the SVP and General Manager for AI Security at Okta, where he leads product, go-to-market, and commercial strategy for securing agentic AI. He has more than 20 years of experience across engineering, product management, marketing, and general management, spanning financial services, technology, and human capital management, with prior roles at Salesforce, ADP, and Proxyclick. He holds an MBA from the Haas School of Business at UC Berkeley. Transcript The Future of AI Security: The Right Architecture for Agents Secure Your Business Against AI Agents

Legendary entrepreneur and investor Mitch Kapor draws on his decades of experience to argue that while AI represents a massive wave of disruptive innovation, it also represents an opportunity to avoid mistakes made with social media and the early internet. In this episode, he contends that technologists tend toward over-optimism about technology solving human problems while underestimating downsides. Self-regulation by large AI companies like OpenAI and Anthropic is likely to fail, he suggests, because incentives to aggregate power and wealth are too strong, requiring external pressure and oversight. Kapor explains that his responsible investing approach at his venture capital firm, Kapor Capital, focuses on gap-closing rather than diversity for its own sake, funding startups that address structural inequalities in access, opportunity, or outcomes, regardless of founder demographics. He discusses the Humanity AI initiative and argues that philanthropy needs to develop AI literacy and technical capacity, with some foundations hiring chief technology officers to effectively engage with these issues. He believes targeted interventions can create meaningful change without matching the massive investments of the major AI labs. Kapor expresses hope that a younger generation of leaders in tech and philanthropy can step up to make positive differences, emphasizing that his generation should empower them rather than occupying seats at the table. Mitch Kapor is a pioneering technology entrepreneur, investor, and philanthropist who founded Lotus Development Corporation and created Lotus 1-2-3, the breakthrough spreadsheet software that helped establish the PC software industry in the 1980s. He co-founded the Electronic Frontier Foundation to advocate for digital rights and civil liberties, and later established Kapor Capital with his wife Freada Kapor Klein to invest in startups that close gaps of access, opportunity, and outcome for underrepresented communities. Kapor recently completed a masters degree at the MIT Sloan School focused on gap-closing investing, returning to finish what he started 45 years earlier when he left MIT to pursue his career in Silicon Valley. He serves on the steering committee of Humanity AI, a $500 million initiative to ensure AI benefits society broadly.

Oliver Patel has built a sizeable online following for his social media posts and Substack about enterprise AI governance, using clever acronyms and visual frameworks to distill down insights based on his experience at AstraZeneca, a major global pharmaceutical company. In this episode, he details his career journey from academic theory to government policy and now practical application, and offers insights for those new to the field. He argues that effective enterprise AI governance requires being pragmatic and picking your battles, since the role isn't to stop AI adoption but to enable organizations to adopt it safely and responsibly at speed and scale. He notes that core pillars of modern AI governance, such as AI literacy, risk classification, and maintaining an AI inventory, are incorporated into the EU AI Act and thus essential for compliance. Looking forward, Patel identifies AI democratization—how to govern AI when everyone in the workforce can use and build it—as the biggest hurdle, and offers thougths about how enteprises can respond. Oliver Patel is the Head of Enterprise AI Governance at AstraZeneca. Before moving into the corporate sector, he worked for the UK government as Head of Inbound Data Flows, where he focused on data policy and international data transfers, and was a researcher at University College London. He serves as an IAPP Faculty Member and a member of the OECD's Expert Group on AI Risk. His forthcoming book, Fundamentals of AI Governance, will be released in early 2026. Transcript Enterprise AI Governance Substack Top 10 Challenges for AI Governance Leaders in 2025 (Part 1) Fundamentals of AI Governance book page

Kevin Werbach speaks with Trey Causey about the precarious state of the responsible AI (RAI) field. Causey argues that while the mission is critical, the current organizational structures for many RAI teams are struggling. He highlights a fundamental conflict between business objectives and governance intentions, compounded by the fact that RAI teams' successes (preventing harm) are often invisible, while their failures are highly visible. Causey makes the case that for RAI teams to be effective, they must possess deep technical competence to build solutions and gain credibility with engineering teams. He also explores the idea of "epistemic overreach," where RAI groups have been tasked with an impossibly broad mandate they lack the product-market fit to fulfill. Drawing on his experience in the highly regulated employment sector at Indeed, he details the rigorous, science-based approach his team took to defining and measuring bias, emphasizing the need to move beyond simple heuristics and partner with legal and product teams before analysis even begins. Trey Causey is a data scientist who most recently served as the Head of Responsible AI for Indeed. His background is in computational sociology, where he used natural language processing to answer social questions. Transcript Responsible Ai Is Dying. Long Live Responsible AI

Cameron Kerry, Distinguished Visiting Fellow at the Brookings Institution and former Acting US Secretary of Commerce, joins Kevin Werbach to explore the evolving landscape of AI governance, privacy, and global coordination. Kerry emphasizes the need for agile and networked approaches to AI regulation that reflect the technology's decentralized nature. He argues that effective oversight must be flexible enough to adapt to rapid innovation while grounded in clear baselines that can help organizations and governments learn together. Kerry revisits his long-standing push for comprehensive U.S. privacy legislation, lamenting the near-passage of the 2022 federal privacy bill that was derailed by partisan roadblocks. Despite setbacks, he remains hopeful that bottom-up experimentation and shared best practices can guide responsible AI use, even without sweeping laws. Cameron F. Kerry is the Ann R. and Andrew H. Tisch Distinguished Visiting Fellow in Governance Studies at the Brookings Institution and a global thought leader on privacy, technology, and AI governance. He served as General Counsel and Acting Secretary of the U.S. Department of Commerce, where he led work on privacy frameworks and digital policy. A senior advisor to the Aspen Institute and board member of several policy initiatives, Kerry focuses on building transatlantic and global approaches to digital governance that balance innovation with accountability. Transcript What to Make of the Trump Administration's AI Action Plan (Brookings, July 31, 2025) Network Architecture for Global AI Policy (Brookings, February 10, 2025) How Privacy Legislation Can Help Address AI (Brookings, July 7, 2023)