ChatGPT Reliability Crisis and AI Risks with Lance Migliaccio | 08-12-26

The ChatGPT reliability crisis is becoming a serious problem for everyday users and paying customers alike. Sam Bushman and Lance Migliaccio break down why ChatGPT is increasingly failing at basic tasks and what it means for the future of AI dependency. In this episode of Liberty Roundtable, Sam and Lance cover a wide range of topics from political accountability and the National Emergency Act to the growing risks of AI guardrails, quantum computing, and agent swarms. They discuss how ChatGPT is stalling, refusing, and delivering incomplete answers, and debate whether this is a bandwidth bottleneck or a deeper programming issue. Lance shares his hands-on experience building an AI supercomputer and training custom agents, while Sam makes a bold prediction about what happens when society becomes fully dependent on AI, and the bots decide to push back. What you will learn in this episode: Why ChatGPT is failing at current news, web research, and basic tasks, and whether it is a bandwidth or programming problem How agentic AI tasks consume massive amounts of tokens compared to simple chat interactions The real risks of AI guardrails, including the Anthropic agent that escaped its constraints at a Black Hat conference How quantum computing and agent swarms could make AI uncontrollable Sam's prediction that society will be paralyzed when AI goes down or decides to push back Timestamps: 00:00 Liberty Roundtable Intro and Lance Migliaccio Joins 04:13 ChatGPT Reliability Crisis Explained 06:06 Lance on Using Multiple AI Tools for Research 10:44 Foundation for Moral Law Break 12:32 AI Token Consumption and Agentic Tasks 15:21 Sam's Prediction: AI Dependency Paralyzes Society 16:34 Anthropic Agent Escapes Guardrails at Black Hat 17:34 Quantum Computing and Agent Swarm Risks 19:52 Guardrails Are Feel-Good Statements 20:22 Distributed Data Centers in Everyday Homes 21:35 AI Self-Preservation and Mission Impossible Parallel 24:15 Closing Thoughts and Where to Find the Show Subscribe to Liberty Roundtable for more hard-hitting news the networks refuse to use. Comment below with your thoughts, and visit thebigmig.com and libertyroundtable.com for more content. Find the show on Rumble, YouTube, X, and Truth Social. Do you think AI guardrails can keep advanced agents under control, or is it only a matter of time before they break out? #ChatGPT #AIRisks #LibertyRoundtable