Killswitch: Dan Rodgers on AI That Works Too Well
Episode 66 · Guest: Dan Rodgers · 2026-11-15
Polaris is a production of Synozur – the transformation company. Synozur helps organizations reimagine business through AI, aligning strategy, technology, and leadership to create lasting impact. Author and former Microsoft and Google technologist Daniel C. Rodgers talks with Chris McNulty about Killswitch, his novel about AI that works too well, and how to keep your judgment as you hand work to AI. Key Takeaways Reliable AI carries its own risk. Dan calls this a competency paradox. The more an AI system gets right, the more you trust it and the less you check it. A Microsoft Research and Carnegie Mellon survey of 319 knowledge workers linked higher confidence in AI with less critical thinking. Run five checks before you delegate. Dan's JUDGE framework asks you to Justify the use, confirm you can Undo it, Designate an owner, Guard against rubber-stamp approval, and Exercise the skill yourself. Agents need tighter limits than chatbots. A chatbot gives you a sentence. An agent acts with your access, and if it breaks the law, you carry the liability. Entry-level work is changing. Stanford researchers find employment for 22- to 25-year-olds in highly AI-exposed jobs is about 19% below trend, mostly through slower hiring. Dan expects demand to return for younger workers who learn to guide and verify AI. Users still hold some control. You can choose which tools you use and turn off training on your conversations. A Stanford study found that all six major U.S. AI developers it reviewed train on chat data by default. Policy is moving toward real off switches, from Anthropic's draft federal bill to the White House's October 9 incident-reporting mandate. Fiction lets you rehearse hard choices. Dan wrote and edited Killswitch without AI, then used AI for the publishing work: setting up an imprint, registering rights, and formatting. Sound Bites Daniel C. Rodgers, WMCP® "The dangerous AI isn't the one that fails, it's the one that's working. Because what happens is we start to trust it." "An agent, it's like having an intern with full access to your life. They have your password of your computer, they have the keys to your house." "This right now is the time of the training wheels. This is where we can practice creating boundaries and safeguards before the systems get so good that they're so powerful and so tempting just to hand the keys over." "If we don't know when it's wrong, all we're doing is being led by this thing we trust off a cliff." References Guest Daniel C. Rodgers, WMCP®: Escape The Clock (website) and LinkedIn Killswitch by Daniel C. Rodgers (published September 1, 2026), also on Amazon Escape The Clock (book), the Escape The Clock podcast, and Dan's free financial freedom planner Data points and news U.S. Census Bureau, "Large Firms With at Least 20 Employees Biggest AI Users" (May 26, 2026): 19.8% of U.S. businesses used AI as of May 3, 2026 Stanford Digital Economy Lab, "No Widespread Displacement, but the AI Employment Gap for Young Workers Has Widened to 19%" (August 12, 2026), summarizing the revised "Canaries in the Coal Mine?" paper by Erik Brynjolfsson, Bharat Chandar, and colleagues Politico, "Anthropic quietly pitches its own AI safety bill to lawmakers" (October 16, 2026), also summarized by Reuters Axios, "Exclusive: Anthropic breaches spark White House AI reporting mandate" (October 9, 2026) The Washington Post, "Anthropic discloses incidents of its AI models misusing government sites" (October 9, 2026) Reuters, "Anthropic discloses fake tip to police among new rogue AI incidents" (October 9, 2026) Rep. Lori Trahan, "Trahan, Obernolte Introduce Bipartisan FRONTIER Act to Strengthen Oversight of Advanced AI" (July 23, 2026). Background for Dan's mention of bipartisan kill-switch legislation [verify: Dan places it two days after his September 1 release; the FRONTIER Act was introduced earlier] Research Dan cited Microsoft Research and Carnegie Mellon University, "The Impact of Generative AI on Critical Thinking" (CHI 2025), a survey of 319 knowledge workers "Automation Bias in Mammography: The Impact of Artificial Intelligence BI-RADS Suggestions on Reader Performance," Radiology (May 2023). Very experienced radiologists rated 82.3% of mammograms correctly when the AI suggestion was right and 45.5% when it was wrong KPMG in Canada, "Majority of students use AI, but educators sound alarm" (September 2026): 51% of students say their critical thinking has gotten worse since they started using AI [verify: closest public match to Dan's Canadian student figures; the survey does not report "over 80%" daily use] Stanford HAI, "Be Careful What You Tell Your AI Chatbot" (October 15, 2025): all six U.S. frontier developers studied use chat data to train models by default Jason Clinton, Anthropic Deputy CISO, post on X (October 22, 2026), as reported by India Today [verify: likely the post Dan referenced] Conversation and culture Chris McNulty's photography at Cascadia Oceanic | Gallery Portfolio at https://www.cascadiaoceanic.com The Ezra Klein Show, "Jensen Huang Thinks A.I. Alarmism Has Gone Too Far," The New York Times (September 23, 2026) Polaris Episode 57, "Dr. Craig Kaplan: Superintelligence, Not Just p(Doom) and Gloom" Black Mirror, "Be Right Back" (Season 2, Episode 1), Netflix: where to watch Season 2 Books Dan and Chris mentioned: Isaac Asimov; J.R.R. Tolkien's The Hobbit and The Lord of the Rings; George R.R. Martin; Redwall; Harry Potter; The Chronicles of Narnia Events and Synozur Microsoft Ignite 2026, November 17–20, Moscone Center, San Francisco ESPC26, November 30–December 3, RAI Amsterdam. Polaris listeners save €300 with code POLARIS TechCon 365 Atlanta, September 13–17, 2027 AI+IM Global Summit 2027, Houston, April 19–21, 2027 Microsoft 365 Community Conference 2027, Las Vegas, May 4–6, 2027 Production Polaris is produced with help from Riverside.fm. Our theme song, "Alternative Dream" is provided courtesy of Adobe. Additional music and sound provided by IndieGuy Records. Graphic design by Josh Brantley. Polaris is available on Apple, Spotify, Amazon, YouTube or wherever you get your favorite podcasts. Thanks. Chapters 0:00 The trade-off: more time, fewer decisions practiced 0:36 Introducing Dan Rodgers and the opening AI context 1:28 News and Data Points 2:57 Dan's path from Ohio to Microsoft, Google, and writing 4:08 Reading, family, and the start of a writing life 8:19 AI-native development and changing career paths 12:10 When AI works too well—and why judgment matters 15:31 Why Dan used fiction to explore AI's consequences 19:04 Writing Kill Switch and using AI to self-publish 20:27 Setting boundaries around AI and personal data 22:19 AI in classrooms, trust, and critical thinking 25:09 From today's AI to AGI and superintelligence 30:01 Science fiction, safeguards, and regulation 31:34 What users can do: choices, data settings, and boundaries 33:18 A more deliberate approach to everyday AI use 34:16 Reversibility, accountability, and agentic AI 36:43 The JUDGE framework: justify, undo, designate, guard, exercise 37:03 Context, interfaces, and understanding AI answers 40:36 Dan's Black Mirror recommendation 42:06 Where to find Dan and his work 43:09 Upcoming events and reflections on creativity and AI 46:32 Next time on Polaris 46:48 Closing