The Other AI: Audio Briefings on Augmented Intelligence and AI Governance

Basil C. Puglisi

The Other AI turns Basil C. Puglisi's articles, white papers, and policy briefs into audio briefings on AI governance, augmented intelligence, human judgment, and human-AI collaboration. The format is built for the time and conditions in which people actually learn, whether running, driving, riding a train, or working on something else. Episodes are AI-narrated for clean, consistent production, and human review approves each publication before release. The complete original work, including details, sources, and citations, lives at basilpuglisi.com. Topics include HAIA-RECCLIN, Factics, Checkpoint-Based Governance, enterprise AI adoption, AI policy, cognitive enhancement, and the future of human authority over automated systems. This podcast is for executives, researchers, consultants, educators, policy thinkers, and AI practitioners who want more than AI hype. The show focuses on evidence, dissent, governance, measurable outcomes, and the role of human judgment when machines become more capable.

  1. Did AI Write the Pope's Encyclical? Scanner Data, Governance, and What Nobody Is Acting On

    1d ago

    Did AI Write the Pope's Encyclical? Scanner Data, Governance, and What Nobody Is Acting On

    Did Pope Leo XIV plagiarize? Did the Pope use AI to construct Magnifica Humanitas? Those were the first two questions when I started reading the encyclical. I ran every chapter through Originality.ai. Every one flagged. Plagiarism scores between 79% and 96%. Chapter Three, the chapter on technology, scored 72% likely AI written. Both readings were wrong. The plagiarism flags came from republications of the encyclical across the internet after its release. The scanner was matching the Pope's words against copies of the Pope's words. The AI detection scores came from topic overlap, not authorship. In this episode, the hosts unpack what happened when an AI governance practitioner ran a papal encyclical through an AI scanner and discovered the tools are functionally useless without human oversight. They dig into why AI detection fails, what it means that a practitioner and a Pope arrived at the same governance diagnosis from opposite directions, and why the entire field can name the problem but nobody seems able to act on it. The conversation covers the scanner failure, peer-reviewed research confirming AI detection tools do not work reliably, the convergence between secular governance frameworks and Catholic moral teaching, the authorship question, and the call for Augmented Intelligence: human governance at pace, not a pause, not a slowdown, not unchecked adoption. Based on the article by Basil C. Puglisi, MPA. Full article: https://basilpuglisi.com/did-ai-write-magnifica-humanitas/ Read the original papers at basilpuglisi.com Spotify: https://open.spotify.com/show/033dvhzMIcWLdY7IUgsu7F Apple Podcasts: https://podcasts.apple.com/us/podcast/id1896506152 Amazon Music: https://music.amazon.com/podcasts/923d1a79-533f-4623-bae3-e2ba83453dfb YouTube: https://www.youtube.com/playlist?list=PLchpU2bIYoEEBh2hdY-BVP9ckyTPiHOAQ This episode was AI generated under NotebookLM as a Deep Dive audio overview, not a polished product. It may contain errors. The full article at the link above is the canonical source. #AIassisted using HAIA Ecosystem

    20 min
  2. The AI Synthesizer Accountability Gap: What Perplexity's Model Council Needs

    3d ago

    The AI Synthesizer Accountability Gap: What Perplexity's Model Council Needs

    Here are all the fields ready for paste: Episode Title: The AI Synthesizer Accountability Gap: What Perplexity's Model Council Needs Episode Notes (paste into the 4,000-char field): Two hosts unpack a governance case study on Perplexity's Model Council, the multi-model AI feature that dispatches Claude, GPT, and Gemini simultaneously and synthesizes their answers into one response. The central question: when an automated synthesizer combines three frontier model outputs before any human reads them, who governs what it includes, excludes, and changes? The conversation covers five areas. First, why parallel multi-model dispatch is the correct architecture for consequential decisions and what Perplexity got right. Second, the synthesizer accountability gap, including seven documented failure modes for AI synthesizers operating without human checkpoint authority. Third, the Bumblebee contrast, where Perplexity's own open-source supply chain scanner already follows a checkpoint workflow (AI drafts, human reviews, then the output ships) but Model Council's synthesizer has no equivalent gate. Fourth, the open-source governance overlay designed to close the gap: structured reasoning at every model, non-cognitive audit infrastructure, and checkpoint-based governance with named human authority. Fifth, the federal stakes, given Perplexity's direct government partnership through GSA's Multiple Award Schedule. Full case study: https://basilpuglisi.com/perplexity-model-council-governance/ Open-source governance spec: https://github.com/basilpuglisi/HAIA Book: Governing AI: When Capability Exceeds Control (ISBN 9798349677687) Basil C. Puglisi, MPA A Human-AI Collaboration basilpuglisi.com Podcast "The Other AI": Spotify: https://open.spotify.com/show/033dvhzMIcWLdY7IUgsu7F Apple: https://podcasts.apple.com/us/podcast/id1896506152 Amazon Music: https://music.amazon.com/podcasts/923d1a79-533f-4623-bae3-e2ba83453dfb YouTube Playlist: https://www.youtube.com/playlist?list=PLchpU2bIYoEEBh2hdY-BVP9ckyTPiHOAQ these are AI generated under NotebookLM as audio overviews not polished products

    21 min
  3. Your Insurance Policy Probably Excludes AI: The Actuarial Gap Nobody Proved

    5d ago

    Your Insurance Policy Probably Excludes AI: The Actuarial Gap Nobody Proved

    No published actuarial study proves that AI governance lowers insurance costs. Seven carriers have filed exclusion endorsements that may have already changed your coverage. The insurance market appears to be sorting organizations by governance evidence without the loss data to validate the sorting. The market moved before the math arrived. This episode explores The AI Risk Economy: Why Insurance Cannot Price What Governance Cannot Prove, a working paper by Basil C. Puglisi, MPA. The conversation covers the Economic Override (the author's framing of the structural tendency for economic incentives to override governance), the proposed five-tier insurance maturity model, the actuarial gap at the center of the emerging practice, and the six questions every organization should answer before their next renewal. The five-tier model and six-variable governance declaration are the author's proposed analytical instruments, not adopted industry standards. The paper documents early market signals and names an instability. It does not claim to resolve it. Key topics in this episode: The dangerous disconnect between coverage assumptions and carrier filings. The five-tier model as one possible lens, from Tier 1 (no AI policy, potentially excluded) to Tier 5 (structured audit records, strongest insurance profile). The Tier 3 to Tier 4 boundary where technical controls end and named human accountability begins. The cyber insurance precedent and the three structural differences that make AI unpredictable. The six-variable governance declaration as a starting point for the renewal conversation. Read the full working paper: basilpuglisi.com/ai-risk-economy-insurance-governance SSRN Abstract ID 6823580 (pending) GitHub: github.com/basilpuglisi/Public-Policy Listen to The Other AI: Audio Briefings on Augmented Intelligence and AI Governance Spotify: https://open.spotify.com/show/033dvhzMIcWLdY7IUgsu7F Apple Podcasts: https://podcasts.apple.com/us/podcast/id1896506152 Amazon Music: https://music.amazon.com/podcasts/923d1a79-533f-4623-bae3-e2ba83453dfb YouTube Playlist: https://www.youtube.com/playlist?list=PLchpU2bIYoEEBh2hdY-BVP9ckyTPiHOAQ This episode is AI generated under NotebookLM as an audio overview, not a polished product. #AIassisted using HAIA Ecosystem

    45 min
  4. The Protocol That Makes AI Oversight Auditable and Insurable

    May 27

    The Protocol That Makes AI Oversight Auditable and Insurable

    Most organizations deploying AI cannot answer the questions their boards, regulators, and insurers are already asking. Can you show what the system actually did? Is the evidence tamper-evident? Does every action trace back to a named human who made a binding decision? Can a third party audit it later? The first market response to this gap was not governance pricing. It was exclusion. Between June 2024 and January 2026, carriers began issuing flat AI exclusions across commercial liability products because insurers cannot price what they cannot observe. This deep dive conversation covers HAIA-CARCS v1.4: Compliance Accountability Record and Case Study, a structured documentation protocol that transforms scattered AI platform histories into verifiable proof of human oversight. The conversation walks through: Why CARCS was created: the EU AI Act Article 14 human oversight mandate, the constitutional need inside Checkpoint-Based Governance for documented proof of engagement, and the set of questions no organization can currently answer. How CARCS works as a single-practitioner method: one person, three prompts, ten sections, four classified decision types. The distinction between Corrective Override and Checkpoint Confirmation is what catches the difference between a human who actually governed and a human who rubber-stamped. How CARCS scales with GOPEL: the Governance Orchestrator Policy Enforcement Layer chains every record using SHA-256 cryptographic hashes, producing tamper-evident, append-only locked audit records. Attestation Grade becomes Hash Verified. Why locked records are still not enough without VAISA: the Verified AI Inference Standards Act closes the inference boundary where data protection law goes blind, requiring hardware attestation to prove data was protected during AI processing. How insurance prices the three-tier framework: Ethical AI produces principles. Responsible AI produces engineering controls. AI Governance produces a named human who answers personally when the system fails. The insurance market is now pricing the difference, and CARCS is the protocol that makes the difference visible. How CARCS v1.4 answers all eleven questions the industry is asking about proof and verification audits for AI use through the Governance Infrastructure Declaration and the published HAIA Ecosystem. Read the full CARCS v1.4 specification: https://basilpuglisi.com/haia-carcs-compliance-accountability-record-case-study/ github.com/basilpuglisi/HAIA Book: Governing AI: When Capability Exceeds Control by Basil C. Puglisi (ISBN 9798349677687) Listen on The Other AI: Audio Briefings on Augmented Intelligence and AI Governance: Spotify: https://open.spotify.com/show/033dvhzMIcWLdY7IUgsu7F Apple Podcasts: https://podcasts.apple.com/podcast/id1896506152 Amazon Music: https://music.amazon.com/podcasts/923d1a79-533f-4623-bae3-e2ba83453dfb YouTube Playlist: https://www.youtube.com/playlist?list=PLchpU2bIYoEEBh2hdY-BVP9ckyTPiHOAQ These are AI generated under NotebookLM as video overviews not polished products. #AIassisted using the HAIA Ecosystem

    24 min
  5. Why AI Needs Infrastructure, Not Regulation: The Congressional Blueprint for Provider Plurality

    May 25

    Why AI Needs Infrastructure, Not Regulation: The Congressional Blueprint for Provider Plurality

    Geoffrey Hinton, widely recognized as one of the godfathers of deep learning, has publicly estimated a 10 to 20 percent probability that artificial intelligence could displace humanity entirely. If a structural engineer told you there was a 20 percent chance the bridge you were about to cross would collapse, you would not drive across it. This episode unpacks a completely different approach to that risk. For years the AI debate has been trapped in a false choice: either let tech companies run unchecked to maintain global competitiveness, or regulate them heavily and risk slowing progress. The congressional package examined here proposes a third path. The answer is not regulation as we traditionally understand it. The answer is infrastructure. The episode covers five core arguments drawn from the AI Provider Plurality legislative package and the Verified AI Inference Standards Act (VAISA), both submitted to the 119th Congress in February 2026 and published open source. The Cognitive Cartel. A handful of corporations control the underlying AI infrastructure of America. That concentration creates a flawed oracle problem and a single point of failure across healthcare, finance, education, and national security. Infrastructure, Not Content Regulation. The historical pattern is clear. Commercial aviation got the FAA in 1958. Finance got the SEC in 1934. Telecommunications got the FCC. In every case, the government built structural infrastructure rather than dictating what companies could say or build. AI requires the same treatment. GOPEL: The Non-Cognitive Highway. The Governance Orchestrator Policy Enforcement Layer is designed as a strictly deterministic, non-cognitive system. It dispatches, collects, routes, logs, pauses, hashes, and reports. It does not evaluate AI outputs. A mail carrier for AI decisions, not a judge. That design is a deliberate security feature: you cannot manipulate a system that has no cognition to exploit. The Invisible Moment and VAISA. Every time a hospital, bank, or school sends sensitive data to an external AI platform for processing, that data enters a window no current law governs. VAISA addresses this gap through a four-tier classification system (Profiles 0 through 3) using hardware-enforced trusted execution environments and cryptographic attestation, modeled on the aviation transponder principle. Provider Plurality and Small Model Investment. An operational test using nine independent AI platforms found that eight converged on the same wrong answer. The ninth dissented and was the only one correct. Without structural diversity in AI systems, there is no comparison point to catch failure. The package calls on Congress to fund small AI platforms through existing SBIR and STTR grant mechanisms to ensure competitive alternatives exist on the road. The legislative strategy requires zero new appropriations to begin. Federal agencies can pilot pluralistic governance workflows immediately using existing statutory authority and current budgets. Source material: AI Provider Plurality Congressional Package (5 documents), Verified AI Inference Standards Act (VAISA), GOPEL v1.5 Canonical Specification, and Governing AI: When Capability Exceeds Control by Basil C. Puglisi (ISBN 9798349677687). All source documents are published open source at github.com/basilpuglisi/HAIA. This episode is AI generated under NotebookLM as an audio briefing, not a polished production. #AIassisted using the HAIA Ecosystem

    47 min
  6. The Evocative Audit: What Metrics Cannot Carry in AI Bias | Deep Dive

    May 23

    The Evocative Audit: What Metrics Cannot Carry in AI Bias | Deep Dive

    What happens when a structured multi-AI review process spends months scanning a book manuscript for accuracy and still misses a foundational concept? Three sentences from Dr. Timnit Gebru on LinkedIn caught what eleven platforms could not. This episode is a deep dive into Dr. Joy Buolamwini's 2022 MIT PhD thesis, "Facing the Coded Gaze with Evocative Audits and Algorithmic Audits," and the concept at its center: the evocative audit. Buolamwini argues that the standard approach to AI accountability, testing systems and publishing error rates, produces evidence that moves institutions but fails to move people. The evocative audit is her formal answer to that gap. It combines human experience with documented evidence using a mechanism she calls the counter-demo to make algorithmic harm visible in ways that numbers alone cannot. The episode covers: What the evocative audit is and how it differs from a standard algorithmic audit The counter-demo mechanism and how it uses a system's own outputs against its claims of accuracy The four types of evocative audits: comparative, participatory, performative, and singular Why Buolamwini grounds her work in Black feminist epistemology and what that means for governance The Gender Shades study and the "AI, Ain't I A Woman?" spoken word performance as paired evidence How the combination produced the 2020 corporate moratoria from IBM, Amazon, and Microsoft Why Dr. Gebru's LinkedIn comment caught a gap that months of multi-AI review did not The historical roots of the counter-demo in the work of Frederick Douglass and Sojourner Truth This episode is based on the article by Basil C. Puglisi: https://basilpuglisi.com/the-evocative-audit-what-metrics-cannot-carry-in-ai-bais/ Dr. Buolamwini's original thesis: https://dspace.mit.edu/entities/publication/4d7bdc57-b375-45c5-aa4b-cd7076a7ebce Dr. Buolamwini's book Unmasking AI (Random House, 2023): https://www.unmasking.ai/ This episode is AI generated under NotebookLM as an audio overview, not a polished product. The AI narration contains known errors including mispronunciation of Dr. Buolamwini's name, Dr. Gebru's name, and the term "excoded" (Buolamwini's coinage for people harmed by algorithmic systems). These are AI generation artifacts that cannot be corrected without regenerating the audio. #AIassisted using the HAIA Ecosystem

    21 min
  7. Replacement or Augmentation: The Tale of Two AIs Deep Dive

    May 21

    Replacement or Augmentation: The Tale of Two AIs Deep Dive

    Two AI deployment architectures are operationally available in 2026. Both are chosen today by named executives at named institutions. They produce opposite consequences for workers, organizations, and society. The choice between them is being made every quarter, in every deployment decision, by named people in named roles, and most professionals making the choice have never seen the evidence laid out. This deep dive walks through that evidence. You will hear how Oracle's $156 billion AI infrastructure commitment was funded by removing thousands from payroll, with Larry Ellison's framing on record: "We are choosing the chips." You will hear how Cynergy Bank deployed the same class of frontier model in the same year and produced 18 percent fewer customer complaints with a workforce that grew rather than shrank. Same technology. Opposite architecture. Different outcomes. You will hear the headcount evidence at scale: 154,445 technology-sector layoff announcements in 2025, the Stanford Digital Economy Lab finding of 16 percent relative employment decline for workers aged 22 to 25 in AI-exposed occupations, and the SignalFire data showing a 50 percent drop in new graduate hiring at the 15 largest technology firms since 2019. You will hear what augmentation looks like at maximum institutional scale. JPMorgan Chase committed $18 billion to technology in 2025 and trained 300,000 workers. Walmart committed $1 billion to skills development reaching 1.7 million US and Canada associates. Noy and Zhang's Science study documented 40 percent task time reduction with 18 percent quality improvement under augmentation deployment. PwC measured a 56 percent wage premium for workers with AI skills. You will hear the Mirror Diagnostic. The Anthropic February 2026 sequence is the most precisely recorded case of deployment incentives overriding stated commitments. The voluntary pause commitment in the Responsible Scaling Policy was removed under competitive pressure. The autonomous weapons and surveillance commitments, structured as public legal redlines, survived the same week at $200 million in lost Pentagon contract revenue. Voluntary commitments fold. Structural commitments hold. The architecture of the commitment is what determines the outcome. You will hear why the third option, prohibition through binding regulation, does not coalesce against globally distributed capability. The EU AI Act's implementation delays, confirmed by the May 2026 Digital Omnibus package extending high-risk obligations to December 2027 and August 2028, are the contemporary instance of a pattern that repeats across alcohol prohibition, drug control, and the 1990s Crypto Wars. You will hear the governance principles that determine which architecture an institution actually operates regardless of which architecture it claims to operate. Named human accountability at consequential decision points. Measurement of cognitive development rather than output volume. Structural commitment survival under economic pressure. This is the evidence base behind the decisions executives, policymakers, and workers face right now. If you make AI deployment decisions, fund them, regulate them, work under them, or simply want to understand them, the working paper is the source you have been missing. The deep dive is the orientation. The paper is the destination. The full working paper "The Tale of Two AIs: Artificial and Augmented" by Basil C. Puglisi, MPA is published at BasilPuglisi.com. The literary novel adaptation publishes this fall 2026. The middle grade companion follows. This podcast is an AI-generated deep dive produced through NotebookLM. The hosts are AI agents. Every claim, citation, and case study discussed is drawn from the working paper, which remains the authoritative source for reference.

    23 min
  8. Governing AI: When Capability Exceeds Control (2025 Book Audio Briefing)

    May 19

    Governing AI: When Capability Exceeds Control (2025 Book Audio Briefing)

    Geoffrey Hinton resigned from Google in 2023 to sound the alarm. He estimates a 10 to 20 percent probability that artificial intelligence causes human extinction within the next 30 years. That warning lands against an enterprise adoption gap already in motion: 76 percent of organizations are actively deploying agentic AI systems, and only 33 percent maintain any responsible AI controls. This 22-minute audio briefing covers the core themes, risks, and operational frameworks from Governing AI: When Capability Exceeds Control by Basil C. Puglisi. The book argues that institutions failing to manage today's operational AI lack the capacity to govern future superintelligence. If authentication tools running 15 to 25 percent false positive rates cannot be fixed today, the same institutions will not validate alignment for systems many orders of magnitude more capable. THE CORE PARADOX: TEMPORAL INSEPARABILITY Present-day failures and future catastrophic risks are linked by institutional capacity, not by technology level. Operational governance built today is the only proven path to civilization-scale safety later. THE ROOT CAUSE: ECONOMIC OVERRIDE Voluntary compliance collapses under competitive market pressure. Roughly 80 percent of frontier compute is controlled by four hyperscale firms, and development concentrates in five companies. Funding asymmetry reinforces the pattern: 98 percent of research investment pursues capability, 2 percent pursues safety and alignment. NINE RISK DOMAINS Corporate concentration and cultural monoculture, echo chambers and polarization, mass surveillance, AI fraud and deepfake disinformation (the 25 million dollar Hong Kong video call scam), biosecurity threats (40,000 toxic VX-like compounds generated in under six hours), autonomous weapons, climate costs of model training, superintelligence acceleration, and operational governance failures. THREE OPERATIONAL FRAMEWORKS Factics Methodology pairs every fact with a tactic and a measurable KPI, converting governance from principle into execution. HAIA-RECCLIN distributes authority across seven specialized roles (Researcher, Editor, Coder, Calculator, Liaison, Ideator, Navigator) so no single perspective dominates and human judgment remains sovereign. Checkpoint-Based Governance establishes mandatory human arbitration at consequential decision junctures and formally documents dissent rather than manufacturing artificial consensus. Preserved Dissent protects valid safety warnings from being buried by executive pressure to launch. THE BOOK AS PROOF OF CONCEPT Governing AI was produced in six weeks through governed collaboration with five AI platforms (ChatGPT, Claude, Gemini, Grok, Perplexity). Across that production, 28 major checkpoint decisions were logged and 26 dissenting opinions were formally preserved by a human arbiter. The empirical results: 96 percent checkpoint utilization and zero hallucination drift over hundreds of pages. The book is the verification of the architecture it describes. THE PROOF STANDARD PROBLEM Civilization-scale catastrophes cannot be the trigger for governance. Pattern-based justification, learning from today's testable operational failures, builds institutional muscle memory before existential stakes arrive. GET THE BOOK Available in eBook and Print on Amazon and Barnes & Noble. ISBN 9798349677687. https://basilpuglisi.com/governing-ai-when-capability-exceeds-control/ Author: https://basilpuglisi.com these are AI generated under NotebookLM as audio overviews not polished products #AIassisted using the HAIA Ecosystem

    22 min
See All (13)

About

The Other AI turns Basil C. Puglisi's articles, white papers, and policy briefs into audio briefings on AI governance, augmented intelligence, human judgment, and human-AI collaboration. The format is built for the time and conditions in which people actually learn, whether running, driving, riding a train, or working on something else. Episodes are AI-narrated for clean, consistent production, and human review approves each publication before release. The complete original work, including details, sources, and citations, lives at basilpuglisi.com. Topics include HAIA-RECCLIN, Factics, Checkpoint-Based Governance, enterprise AI adoption, AI policy, cognitive enhancement, and the future of human authority over automated systems. This podcast is for executives, researchers, consultants, educators, policy thinkers, and AI practitioners who want more than AI hype. The show focuses on evidence, dissent, governance, measurable outcomes, and the role of human judgment when machines become more capable.

Information