Ai Change Desk | Podcast in RSS.com

Ai Change Desk

di Michael Hanna-Butros Meyering

Stagione 1

AI Change Desk | EP040: When a Prompt Becomes a File

A long paste can become an attachment, and Voice can now work with files and Project context. EP040 gives operators a six-part default-state receipt.

AI Change Desk | EP039: Whose Account Did the Agent Use?

AI CHANGE DESK | EP039: WHOSE ACCOUNT DID THE AGENT USE? EPISODE SUMMARY An employee asks an AI agent to send a file. The employee is allowed to run the agent, the connector accepts the request, and every dashboard turns green. But the connector authenticates with the account of the person who built the agent six months ago. Whose authority actually moved the work? This episode extends the receipt framework from episodes thirty-seven and thirty-eight. Michael separates audience permission, credential capability, organizational purpose, and action approval; explains why disclosure is necessary but incomplete; and introduces a paired authority-and-privacy receipt for connected agent workflows. The operating principle is simple: the agent has a name, but the credential carries the authority. A useful audit trail must preserve both. WHAT CHANGED • OpenAI's current Workspace Agents guidance makes the risk of publishing agents with personal connections explicit: other authorized users may be able to act through the creator's authenticated connection. • European Commission guidance says Article 50 transparency obligations under the EU AI Act began applying on August 2, 2026, with duties depending on role, context, system type, and applicable exceptions. • Microsoft guidance recommends dedicated agent identities, named owners and approvers, effective-permission review, correlation identifiers, on-behalf-of-user evidence, and tested revocation. • GitHub's agentic audit fields provide a platform-specific example of separating the agent, session, action, and initiating user. • OpenAI's Health documentation illustrates why disconnecting a source, deleting synced data, and deleting conversation history are separate privacy events. WHAT THIS MEANS FOR OPERATORS • Permission to run an agent is not authority to use every credential connected to it. • Record the requester, agent owner, publisher, approved audience, trigger, session, connection owner, authenticating account, effective downstream scope, action, approval, defender event, and final disposition. • Keep audience permission, credential capability, purpose authority, and action approval as separate decisions. Do not average them into one green status. • Place a data-handling receipt beside the authority receipt: purpose, minimum data needed, actual data returned, recipient, onward sharing, memory, retention, deletion, and required disclosure. • Test revocation. Disable the agent, rotate or remove a credential, invalidate the old token, and prove the old path no longer works. • Treat vendor documentation as a control map, not proof of your tenant's configuration or runtime behavior. THIS WEEK'S 45-MINUTE BLOCK Choose one connected AI workflow that can retrieve data or take an action. 1. Spend ten minutes mapping the requester, agent owner, publisher, approved audience, trigger, agent/session fields, connection owner, and authenticating account. 2. Spend ten minutes recording the effective downstream scope. Separate read, write, send, share, schedule, edit, and delete. Record which actions require approval. 3. Spend ten minutes mapping purpose, data category, minimum needed, actual data returned, recipient, onward sharing, memory, retention, deletion, and disclosure. 4. Spend ten minutes running one allowed action and one denied action. Remove or rotate one connection and prove the old path no longer works. Capture both agent-side and defender-side evidence. 5. Spend five minutes reconciling identities, timestamps, purpose, data returned, approval, and revocation. Record every mismatch, owner, correction, residual risk, and final disposition. Keep the workflow supervised until the receipts reconcile. LISTENER QUESTION Can your team prove which account sup...

AI Change Desk | EP032: Memory Summary Exit Check

If AI memory can be edited more visibly, turned off more easily, and still be rebuilt from old context later, the operating question is not whether the settings page looks cleaner. It is what evidence proves sensitive or stale context actually left the workflow. Why OpenAI's June 12 memory-summary controls matter operationally. Why deleting visible memories is not the same as deleting past chats or every source. Why a partial memory summary creates a partial-ledger problem. How Developer mode and model retirement add inspection and version-receipt pressure. A forty-five minute Memory Summary Exit Check operators can run this week. OpenAI ChatGPT release notes: https://help.openai.com/en/articles/6825453-chatgpt-release-notes OpenAI Memory FAQ: https://help.openai.com/en/articles/8590148-memory-faq/ OpenAI Codex browser docs: https://developers.openai.com/codex/app/browser OpenAI Lockdown Mode: https://help.openai.com/en/articles/20001061-lockdown-mode OpenAI memory product post: https://openai.com/index/chatgpt-memory-dreaming/ YouTube AI labels update: https://blog.youtube/news-and-events/improving-ai-labels-viewers-creators/ Podnews AI disclosure guidance: https://podnews.net/update/ai-disclosures AI-assisted tools were used in parts of the research and production workflow. Final editorial judgment, risk posture, and release approval stay human-led. This is operational guidance, not legal advice.

AI Brief | EP008: Model release control validation

Two current operator signals, translated into a plain-language weekly control block. OpenAI announced plans to acquire Promptfoo, pushing testing/eval workflows further into default AI release practice. Anthropic launched The Anthropic Institute while NIST reinforced monitoring guidance context for deployed AI systems. A 35-minute operator block you can run weekly with one owner and clear pause authority. Require a tiny evidence packet for each AI behavior change (3 prompts + pass/fail + approver + rollback owner). Publish a one-page operator memo in plain language (approved, restricted, paused, exception path, next review). Run one mini pause drill each week: "output is wrong; who pauses in 10 minutes?" Block scale-up on any workflow missing named approver or rollback owner. 00:00 Cold open + framing 00:55 Boundary note complete / theme intro in 01:10 Signal 1: OpenAI/Promptfoo and release evidence 03:58 Signal 2: Anthropic Institute + NIST monitoring pressure 06:05 Next-week 35-minute action block 07:25 Close + outro https://openai.com/index/openai-to-acquire-promptfoo/ https://www.promptfoo.dev/blog/promptfoo-joining-openai https://techcrunch.com/2026/03/09/openai-acquires-promptfoo-to-secure-its-ai-agents/ https://www.anthropic.com/news/the-anthropic-institute https://www.theverge.com/ai-artificial-intelligence/892478/anthropic-institute-think-tank-claude-pentagon-jack-clark https://www.nist.gov/news-events/news/2026/03/new-report-challenges-monitoring-deployed-ai-systems https://www.nist.gov/publications/challenges-monitoring-deployed-ai-systems-center-ai-standards-and-innovation Episode page: https://michaelhbm.com/AiChangeDesk/episodes/brief-2026-03-11-ai-brief Apple Podcasts: https://podcasts.apple.com/us/podcast/ai-change-desk/id1876677295 Spotify: https://open.spotify.com/show/5X1sLLTeULqFCdt7aaisGD AI-assisted tools were used in parts of research and production support. Final editorial judgment and release approval remained human-led. This is operational guidance, not legal advice.

AI Change Desk | EP007: Security Workflow Control Contract

AI CHANGE DESK | EP007: SECURITY WORKFLOW CONTROL CONTRACT If your AI can find a vulnerability, draft a patch, and open a PR, your biggest risk is no longer detection quality. Your biggest risk is workflow ownership: • who can analyze, • who can approve, • who can merge, • who can pause, • and who can attest the execution chain under pressure. This episode translates four current signals into one operational playbook for next week. WHAT CHANGED THIS WEEK 1. OpenAI launched Codex Security in research preview (2026-03-06). 2. Anthropic + Mozilla published concrete AI-assisted vulnerability workflow details (2026-03-06), including CVD and exploit-analysis references. 3. NIST published AI 800-4 on monitoring deployed AI systems (2026-03-06). 4. OpenAI launched GPT-5.4 and ChatGPT for Excel beta (2026-03-05), expanding business-user AI execution surfaces. OPERATOR TRANSLATION • Treat AI security pipelines as action-controlled workflows, not assistant features. • Separate discovery throughput from remediation readiness. • Move monitoring from dashboarding to a named ownership control. • Add spreadsheet-AI usage controls where sensitive decisions or data handling occur. MONDAY BLOCK (45 MINUTES, ONE OWNER) • Minute 0-10: action matrix lock (Analyze, Draft fix, Open PR, Merge, Deploy) with allowed/checkpointed/restricted levels. • Minute 10-20: credential and identity check (remove over-scoped inherited credentials). • Minute 20-30: evidence contract (logs, retention, export path, access controls). • Minute 30-40: disclosure + rollback ownership (name owners, define stop authority). • Minute 40-45: operator memo (what changed, what is approved, what is restricted, who approves exceptions, next review date). LINKS • Episode page: https://michaelhbm.com/AiChangeDesk/episodes/ep007-security-workflow-control-contract • YouTube channel: https://www.youtube.com/@AIChangeDesk • RSS show: https://media.rss.com/aichangedesk/feed.xml • Apple Podcasts: https://podcasts.apple.com/us/podcast/ai-change-desk/id1876677295 • Spotify: https://open.spotify.com/show/5X1sLLTeULqFCdt7aaisGD SOURCES • OpenAI (2026-03-06): https://openai.com/index/codex-security-now-in-research-preview/ • Anthropic + Mozilla collaboration post (2026-03-06): https://www.anthropic.com/news/mozilla-firefox-security • Anthropic coordinated disclosure policy (2026-03-06): https://www.anthropic.com/coordinated-vulnerability-disclosure • Anthropic exploit analysis (2026-03-06): https://red.anthropic.com/2026/exploit/ • Mozilla Firefox blog corroboration (2026-03-06): https://blog.mozilla.org/en/firefox/hardening-firefox-anthropic-red-team/ • NIST AI 800-4 publication page (2026-03-06): https://www.nist.gov/publications/challenges-monitoring-deployed-ai-systems-center-ai-standards-and-innovation • OpenAI GPT-5.4 launch (2026-03-05): https://openai.com/index/introducing-gpt-5-4/ • OpenAI ChatGPT for Excel (2026-03-05): https://openai.com/index/chatgpt-for-excel/ DISCLOSURE AI-assisted tools were used in parts of the research and production workflow. Final editorial judgment, risk posture, and release approval stayed human-led. This is operational guidance, not legal advice. These are my opinions and are not representative of any organization.

Episode 06: AI Brief: GPT-5.3 and continuity controls

Two current operator signals, translated into one concrete next-week action block. OpenAI released GPT-5.3 Instant and published system-card details. Vendor continuity pressure stayed elevated through Anthropic policy-dispute and blacklist-risk signals. A 30-minute Monday control loop to keep model release and fallback controls current. Treat model releases as workflow change events, not just product updates. Run a 3-prompt regression pack before broad rollout after model changes. Confirm rollback owner + stop authority for critical AI workflows. Define one tested fallback path for top three AI-enabled workflows. Send a plain-language operator memo each Monday (approved/restricted/escalation). 00:00 Cold open + framing 00:39 Boundary note complete / theme intro in 00:54 Signal 1: GPT-5.3 Instant and release governance 02:25 Signal 2: vendor continuity pressure 03:45 Monday action block (30-minute control loop) 04:31 Close + outro https://openai.com/index/gpt-5-3-instant/ https://openai.com/index/gpt-5-3-instant-system-card/ https://www.anthropic.com/news/statement-comments-secretary-war https://techcrunch.com/2026/03/02/tech-workers-urge-dod-congress-to-withdraw-anthropic-label-as-a-supply-chain-risk/ https://techcrunch.com/2026/02/27/anthropic-vs-the-pentagon-whats-actually-at-stake/ Episode page: https://michaelhbm.com/AiChangeDesk/episodes/brief-2026-03-04-ai-brief Apple Podcasts: https://podcasts.apple.com/us/podcast/ai-change-desk/id1876677295 Spotify: https://open.spotify.com/show/5X1sLLTeULqFCdt7aaisGD AI-assisted tools were used in parts of research and production support. Final editorial judgment and release approval remained human-led. This is operational guidance, not legal advice.

AI Change Desk | EP005: Run Agents Without Losing Control

AI CHANGE DESK | EP005: RUN AGENTS WITHOUT LOSING CONTROL If AI systems can execute actions in your environment, governance has to move from policy language to access control execution. This episode translates current signals into practical controls for operators: action-tier permissions, scoped credentials, human approval thresholds, deployment tier decisions, and a weekly control desk teams can run quickly. WHAT YOU WILL GET • A practical access-control framework for agent-enabled workflows. • Action-tier classification you can apply this week (read, draft, update-internal, external-send, system-admin). • A deployment control checklist for connected/hybrid/disconnected environments. • A standards-aligned procurement starter (identity, interoperability, proportional controls). • A Monday control desk + metrics scorecard + 30-60-90 implementation sequence. TIMESTAMPS • 00:00 Cold open — access control is the operating risk • 00:50 Intro, disclosure, and show contract • 02:15 Why EP005 now (bridge from EP003 + EP004) • 04:10 Story 1 — Anthropic + Vercept and action-tier controls • 08:30 Story 2 — OpenAI elevated-risk controls and malicious-use patterns • 12:10 Story 3 — Sovereign deployment and architecture obligations • 15:35 Story 4 — NIST standards + proportional controls • 18:55 Scenario walkthrough + risk check • 21:40 Monday Access Control Desk • 24:15 Metrics, 30-60-90 plan, FAQ, and control drills • 25:04 Close + outro MONDAY ACTIONS (RUN THIS NEXT WEEK) 1. Classify top five AI workflows by action tier. 2. Scope credentials for the highest-impact workflow. 3. Name stop-authority owner for each critical workflow. 4. Set approval thresholds for external-send and system-admin actions. 5. Publish one-page operator update with approved/restricted actions and escalation path. SOURCES • https://www.anthropic.com/news/anthropic-acquires-vercept • https://techcrunch.com/2026/02/25/anthropic-acquires-vercept-to-expand-computer-use-agents/ • https://openai.com/index/introducing-lockdown-mode-and-elevated-risk-labels-in-chatgpt-safety/ • https://openai.com/index/disrupting-malicious-ai-uses/ • https://www.microsoft.com/en-us/microsoft-cloud/blog/2026/02/24/announcing-sovereign-cloud-ai-updates/ • https://www.microsoft.com/en-us/industry/blog/government/2026/02/24/accelerating-government-mission-with-microsoft-sovereign-cloud/ • https://www.nist.gov/caisi/ai-agent-standards-initiative • https://www.nist.gov/artificial-intelligence/ai-agent-interoperability-and-efficiency-standards-request-information • https://digital-strategy.ec.europa.eu/en/library/eu-ai-office-and-jrc-publish-report-proportionality-ai • https://ai-watch.ec.europa.eu/publications/eu-ai-office-and-jrc-report-proportionality-trustworthy-ai LISTEN • YouTube: https://www.youtube.com/@AIChangeDesk • Spotify: https://open.spotify.com/show/5X1sLLTeULqFCdt7aaisGD • Apple Podcasts: https://podcasts.apple.com/us/podcast/ai-change-desk/id1876677295 LISTENER QUESTION Where is your organization most exposed right now: permission scope, approval thresholds, or action logging? DISCLOSURE AI-assisted tools were used in parts of drafting, synthesis, and production support. Final editorial judgment and release approval remained human-led.

AI Brief: what changed this week

Two operator-relevant signals from this week, translated into concrete controls teams can execute immediately. Distillation attacks moved from model-lab concern to enterprise operations risk. NIST's AI Agent Standards Initiative reinforced near-term interoperability and accountability expectations. A 25-minute weekly governance desk loop you can run every Monday. Treat provider security bulletins as workflow events, not background reading. Classify AI usage into open-assist, controlled-assist, and restricted classes. Add interoperability and control portability checks to AI procurement intake. Require a human accountability map for every agent-like workflow. Ship a one-page operator update: what changed, what to do, what not to do. 00:00 Cold open: policy that cannot survive Monday is policy theater 01:00 Theme intro 01:16 Framing and disclosure 01:57 Signal 1: distillation attacks and model-control hardening 04:30 Signal 2: standards momentum as procurement and controls signal 06:57 Monday checklist: 25-minute governance desk 08:06 Close 08:18 Final reminder: one owner, one decision, one due date 08:27 Brand outro https://www.anthropic.com/news/detecting-and-preventing-distillation-attacks https://www.businessinsider.com/anthropic-deepseek-distillation-minimax-moonshot-ai-2026-2 https://www.nist.gov/caisi/ai-agent-standards-initiative https://www.ansi.org/standards-news/all-news/2-18-26-nist-launches-ai-agent-standards-initiative https://www.nist.gov/news-events/news/2026/02/nist-seeks-public-input-advance-ai-agent-interoperability-and-efficiency Website episode page: https://michaelhbm.com/AiChangeDesk/episodes/brief-2026-02-25-ai-brief Apple Podcasts: https://podcasts.apple.com/us/podcast/ai-change-desk/id1876677295 Spotify: https://open.spotify.com/show/5X1sLLTeULqFCdt7aaisGD AI-assisted tools were used in research and production support. Final editorial judgment and release approval remained human-led.

AI governance implementation for operators: turning policy into weekly execution

EP003: AI GOVERNANCE IMPLEMENTATION FOR OPERATORS AI governance breaks when it lives as a policy document and not as a weekly operating loop. In this main episode, we use current market signals (model updates, AI security tooling, regional deployment strategy, and standards activity) to show how leaders and operators can run governance as execution instead of theory. WHAT YOU WILL GET • A practical model-change governance workflow you can run every week. • Security workflow controls for AI-assisted code review. • Procurement and data-governance actions triggered by regional/partner deployment signals. • A reusable weekly AI Governance Desk format with owner, controls, and communication outputs. • A late-update block on alignment-research funding and regulated-industry deployment signals. TIMESTAMPS • 00:00 Cold open — governance is a workflow, not a PDF • 00:59 Intro music + disclosure • 01:20 Why this episode now (EP001/EP002 bridge) • 03:20 Story 1 — Claude Sonnet 4.6 and model-change governance • 07:50 Story 2 — Claude Code Security and human-in-the-loop controls • 12:20 Story 3 — OpenAI for India + Tata and procurement reality • 16:00 Story 4 — NIST AI agent interoperability signal • 18:10 Late updates — alignment funding + regulated-industry collaboration • 19:00 Weekly AI Governance Desk (25-minute operating loop) • 22:05 Postscript — chat-code controls + workflow-class policy mapping • 23:25 Monday morning actions • 24:25 Outro + listener question MONDAY MORNING ACTIONS 1. Name one owner for weekly AI governance desk operations. 2. Run a model-change regression check on your top workflows. 3. Require human approval for AI-generated security patches/findings. 4. Update procurement clauses (data handling, change notifications, sub-processors). 5. Publish a one-page internal update: what changed, what to do, what not to do. SOURCES • https://www.anthropic.com/news/claude-sonnet-4-6 • https://docs.anthropic.com/en/release-notes/api#feb-17th-2026 • https://www.anthropic.com/news/claude-code-security • https://docs.anthropic.com/en/docs/claude-code/security • https://openai.com/index/openai-for-india/ • https://www.tata.com/newsroom/openai-and-tata-group-announce-strategic-collaboration • https://www.nist.gov/news-events/news/2026/02/nist-seeks-public-input-advance-ai-agent-interoperability-and-efficiency • https://www.federalregister.gov/documents/2026/02/20/2026-02979/ai-agent-interoperability-and-efficiency-standards-request-for-information • https://openai.com/index/advancing-independent-research-ai-alignment/ • https://alignmentproject.aisi.gov.uk/ • https://www.anthropic.com/news/anthropic-infosys • https://www.infosys.com/newsroom/press-releases/2026/advanced-enterprise-ai-solutions-industries.html LISTEN • Spotify: https://open.spotify.com/show/5X1sLLTeULqFCdt7aaisGD • Apple Podcasts: https://podcasts.apple.com/us/podcast/ai-change-desk/id1876677295 DISCLOSURE AI-assisted tools were used in parts of drafting, synthesis, and production support. Final editorial judgment and release approval remained with the host.

AI policy basics for operators: what this week changed

EP002: AI policy basics for operators. This episode translates AI policy concepts into practical operating decisions for leaders, managers, and delivery teams. Episode: 002 Title: AI policy basics for operators Runtime: 10m 30s Host: Michael Hanna-Butros Meyering AI policy works only when it is written as operational guidance people can apply in daily workflows. 00:00 Why AI policy fails in real teams 01:20 Story 1: Claude Sonnet 4.6 and model-change governance 04:40 Story 2: AI infrastructure cost signals and procurement controls 07:40 Action block: policy + change management implementation 09:40 Monday-morning actions + outro Anthropic launched Claude Sonnet 4.6 (February 17, 2026), which reinforces the need for model-upgrade controls and evaluation gates in internal policy. Anthropic announced it will cover electricity price increases tied to data-center growth (February 17, 2026), making infrastructure impact a practical procurement and governance issue. Scope: which AI use cases are allowed, restricted, or prohibited. Data: which data classes may be used with which tools. Controls: review, logging, exception handling, and escalation. Accountability: who owns policy updates and incident response. Add a model-change trigger section to your AI policy (when re-evaluation is mandatory). Add three infrastructure-risk questions to AI vendor intake. Run one manager briefing with a clear script for allowed/restricted use. Audit one active AI workflow for drift between policy and real usage. Anthropic, “Announcing Claude Sonnet 4.6”: https://www.anthropic.com/news/claude-sonnet-4-6 TechCrunch coverage, “Anthropic releases Claude Sonnet 4.6”: https://techcrunch.com/2026/02/17/anthropic-releases-claude-sonnet-4-6/ Anthropic, “Covering electricity price increases from AI data centers”: https://www.anthropic.com/news/covering-electricity-price-increases Reuters coverage (via Investing.com): https://www.investing.com/news/stock-market-news/anthropic-to-cover-electricity-price-increases-in-areas-where-it-builds-data-centers-3894580 NIST AI Risk Management Framework: https://www.nist.gov/itl/ai-risk-management-framework NIST Generative AI Profile: https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence OECD AI Principles: https://oecd.ai/en/ai-principles ISO/IEC 42001 overview: https://www.iso.org/standard/81230.html This episode uses AI-assisted production tools (voice rendering, editing support, and publishing automation). Final editorial and risk decisions are human-led.

1 di 4