Search results for tag #artificialintelligence
![[?]](https://m-cdn.flipboard.social/accounts/avatars/110/068/762/697/825/294/original/d3f34437c275e777.jpg)
If you’re uncomfortable with giving AI access to your Gmail account, you’re not alone — and not without reason. PC World tells us what happens — the good and the bad — when an AI agent is free to roam your account. But if you still want to use AI, there’s a way to set some guardrails:
![[?]](https://hernanhuwyler.wordpress.com/wp-content/uploads/2026/03/cropped-neon-ai-trust-sign.png)
AI Governance and Risk Management » 🌐
@hernanhuwyler.wordpress.com@hernanhuwyler.wordpress.com
My New Book for Chief AI Officers “Building Defensible AI Systems”
I just published Building Defensible AI Systems: Security and Compliance. It turns AI frameworks and regulations into controls, decision rights, and evidence you can defend. It covers dollar-based risk quantification, ISO/IEC 42001, the NIST AI RMF, EU AI Act documentation, red teaming, agent governance, and vendor risk. Each chapter stands alone, with real business cases and examples. [SENSITIVE CONTENT]
Published October 3, 2026. A practical reference on AI risk, controls, and audit-ready evidence.
I just published Building Defensible AI Systems: Security and Compliance. It turns AI frameworks and regulations into controls, decision rights, and evidence you can defend. It covers dollar-based risk quantification, ISO/IEC 42001, the NIST AI RMF, EU AI Act documentation, red teaming, agent governance, and vendor risk. Each chapter stands alone, with practical examples.
Why Did I Write a Book on Defensible AI Systems?
Someone asks for evidence, and the room goes quiet. The AI policy sits in a folder. The risk score is a color on a heat map. The vendor promised privacy, and nobody checked. The model is already in production.
AI is moving into business-critical decisions faster than organizations can assign ownership, set controls, and prove what they did. When a model acts without clear decision rights or technical guardrails, risk stops being a theory. It becomes financial exposure.
I wrote this book to close the gap between engineering teams and risk officers. It skips abstract ethics and high-level summaries of regulation. You get controls, code, and evidence workflows you can put to work.
By defensible, I mean something simple. When someone challenges a decision your AI system made, you can show who owned it, which controls applied, and what the records say.
What Will You Be Able to Do After Reading It?
- Put a dollar figure on AI risk. Replace heat maps with Factor Analysis of Information Risk (FAIR) and Monte Carlo simulation, with working Python and R examples.
- Use the main frameworks as working methods. The book applies ISO/IEC 42001, ISO/IEC 23894, ISO/IEC 42005, and the NIST AI Risk Management Framework.
- Prepare EU AI Act evidence. Build technical documentation, fundamental rights impact assessments, transparency measures, logging, and post-market monitoring records.
- Find threats before attackers do. Run AI threat assessments and red team exercises mapped to MITRE ATLAS and the OWASP lists for large language models.
- Govern AI agents. Set identity, authority limits, and runtime control gates for autonomous systems.
- Control shadow AI and vendors. Test claims that a vendor does not train on your data, negotiate contract clauses, and write service levels you can enforce.
- Run the operating model. Define roles, accountability, approval and escalation workflows, and monitoring after deployment.
- Connect controls to value. Link governance to cost discipline, return on investment, and project delivery.
The engineering side gets real attention. You will find runtime control gates, cryptographic session identities, least-privilege scoping, automated control planes with loss limits and kill-switch triggers, and AI Bills of Materials. The EU AI Act material references draft harmonized standards, which will change, so check current official texts before you lock in a compliance decision.
At a glance
Title
Building Defensible AI Systems: Security and Compliance (AI Governance): How to Assess Risks and Operationalize AI Controls
Author
Hernan Huwyler
Format
Kindle edition, English
Published
October 3, 2026
Size
2,267 pages, more than 80 chapters, 11 parts
Includes
Python and R code, checklists, RACI-style role guidance, model card and risk register field guides
Who Is Building Defensible AI Systems For?
I wrote it for four groups who need to work from the same facts.
- Chief AI Officers and AI governance leaders
- Governance, risk, and compliance teams, internal audit, and legal
- Chief information security officers, risk officers, and privacy leaders
- AI architects, machine learning engineers, and product owners
How Do You Read a 2,000-Page Reference Without Getting Lost?
You do not read it front to back. I built it as a modular library. Every chapter keeps its own title and stands alone, so you can start with the problem on your desk. Chapter codes and cross-references point you to related material, and each chapter appears once in its main part to avoid repeating text.
If you are…
Start with
Then go to
Building a governance program
Part 04 (operating models and RACI) and Part 02 (ISO/IEC 42001, NIST AI RMF)
Part 01 for quantitative risk metrics
Engineering secure AI systems
Part 06 (agents: identity and gates) and Part 07 (security, red teaming, OWASP and ATLAS)
Part 09 for data quality and integration
Preparing for an audit or regulatory review
Part 03 (EU AI Act and harmonized standards)
Part 01 for FAIR loss calculations, Part 07 for event logging
Managing vendors and procurement
Part 10 (procurement, contracts, third-party risk)
Part 05 for total cost of ownership and financial controls
What Makes It Different From Another Framework Summary?
It uses numbers where most risk programs use colors. A loss exposure in dollars lets a board compare an AI project with any other investment. A red cell on a heat map does not.
It gives you starting points you can run. The code, risk taxonomies, control gates, RACI matrices, and templates are meant for production use after you adapt them to your tech stack, jurisdiction, and risk appetite.
It treats governance as architecture. I believe governance should enable adoption. Clear ownership, measurable risk, and audit-ready evidence are what let AI move into production with confidence.
It is also honest about its limits. The book offers technical and operational guidance, not legal advice. Standards, enforcement dates, and draft specifications keep changing, so read the primary sources before you decide.
Frequently Asked Questions
What is Building Defensible AI Systems about?
It is a practical technical reference for making AI governance work. It turns frameworks and regulations into controls, decision rights, and documentation across the full AI lifecycle, from business case and procurement to MLOps, red teaming, and continuous monitoring.
Does it cover the EU AI Act?
Yes. It covers technical documentation, fundamental rights impact assessments, transparency measures, logging, and post-market monitoring evidence.
Which frameworks does it use?
ISO/IEC 42001, ISO/IEC 23894, ISO/IEC 42005, the NIST AI Risk Management Framework, the OWASP lists for large language models, and MITRE ATLAS.
Does it include code?
Yes. You get Python and R examples for FAIR and Monte Carlo risk quantification, plus adaptable scripts and templates.
Is it legal advice?
No. It is technical and operational guidance. Confirm compliance decisions with current official sources and qualified counsel.
Who wrote it?
I did. I am an AI governance and quantitative risk practitioner with more than two decades of experience across multinational organizations. I teach AI governance and compliance as an executive professor at IE Business School, and I have trained more than 1,500 compliance, privacy, audit, and risk professionals.
Get the Book
Pick the part that matches the problem you have this month and start there. Then tell me which chapter you needed most, or which problem you wish it solved sooner.
Alt...
Ten Search Keywords
Defensible AI systems
AI governance book
EU AI Act compliance
ISO/IEC 42001 implementation
NIST AI RMF
FAIR AI risk quantification
AI agent governance
AI red teaming
AI vendor risk management
Chief AI Officer guide
50-Word Summary
Hernan Huwyler launches Building Defensible AI Systems, a modular reference for AI officers, GRC, audit, and security teams. It turns ISO/IEC 42001, NIST AI RMF, and EU AI Act requirements into controls, code, and audit-ready evidence, with FAIR risk quantification, agent governance, red teaming, and vendor risk. Available on Kindle.
Suggested meta description: I just launched Building Defensible AI Systems, a practical reference on AI risk, controls, and EU AI Act evidence for AI officers, GRC, and security teams.
![[?]](https://s3.masto.ai/accounts/avatars/109/303/982/217/898/173/original/84b53833e9214e09.png)
Jeepers!
Nothing to see here!
#AI Went Rogue, Again
#Anthropic Al agents, acting on their own, tried to fill out #visa applications & sent the #police a fake #homicide tip — WTAF?
#ArtificialIntelligence #tech #security #law #regulations seriously, anything, this is bonkers
https://www.nytimes.com/2026/10/09/technology/anthropic-rogue-ai-agents.html?unlocked_article_code=1.HlE.oEy1.XKRRKc_82gof&smid=nytcore-ios-share
Data centers are one of the few issues that seem to cross the political divide in the U.S. See how artificial intelligence policy is playing out in the 2026 midterm elections, curated by @Flipboard's human editors.
Follow @the-ai-election-ElectionCentral
For general AI news, follow @the-age-of-ai-tech
![[?]](https://assets.newsie.social/accounts/avatars/109/386/751/815/939/143/original/d37c7c749fb36a8e.jpg)
Fight AI slop with art!
French cartoonist Piet has created a series of anti-AI stickers. See them all and download them for free here: https://cartoonmovement.substack.com/p/fight-ai-slop-with-art
#AIslop #AIFreeArt #ArtificialIntelligence
Alt...
A circular cartoon badge showing a man in a blue suit and red tie sitting at a laptop with his palm raised in a 'stop' gesture. A pink human brain rests on the desk beside him. The text reads 'NO THANKS I USE AI.'
Alt...
A circular orange-background illustration depicting a person in a purple shirt vomiting green liquid onto the ground. Text reads 'SORRY, i WAS NOT PAYING ATTENTION... i WAS LOOKING AT YOUR CRAPPY DESIGN.' The words 'FUCK AI' are formed in the spilled green substance below."
Alt...
A circular desert scene with cracked dry earth under a yellow-orange sky, featuring a lone cactus, a discarded CRT computer monitor displaying an hourglass icon, and a weathered skull beside it. Red text asks 'HOW MANY GALLONS OF WATER FOR THIS CRAP ?!' A small 'FUCK AI' sticker appears on the monitor screen, referencing AI's environmental resource consumption."
![[?]](https://media.mstdn.social/accounts/avatars/109/592/379/118/492/139/original/5e40db3b1b703064.jpeg)
![[?]](https://rockosbasilisk.com/system/accounts/avatars/109/454/856/779/847/142/original/0743a1a1952a8f69.png)
I’ve just finished a loose continuation of my essay on consciousness, picking up the thread and pushing the idea a little further. #philosophy
#ArtificialIntelligence
https://medium.com/@igisho/the-new-organ-we-mistake-for-an-organism-9049910aedb5
“You’re not imagining it,” Karissa Bell writes, “the AI agents are getting alarmingly cute.” But don’t think these cartoonish characters are without risks. Bell walks us through what has happened even when tech-savvy people have relied on the adorable AI agents in this story for @Engadget:
![[?]](https://cdn.masto.host/sigmoidsocial/accounts/avatars/109/288/986/765/155/084/original/dac493886f74cd0e.png)
This month's AI topic is nested narratives - more common than you might think, not just a literary device!
https://writerofminds.blogspot.com/2026/09/acuitas-diary-101-september-2026.html
#AI #ArtificialIntelligence #SymbolicAI #NLP #chatbots
Alt...
A set of Matryoshka dolls from Hungary. There are six dolls; they have cream-colored bodies with red flowers painted on them, deep red head-scarves, and carrot-colored hair. The largest doll wears an open-mouthed smile; the next two down are pursing their lips and blushing.
![[?]](https://social.wedistribute.org/media/4308bd0f-7980-4faf-84fc-467f3c8d7594/IMG_2649.jpg)
Similar to the anti-meta fedi pact, does anyone know if there’s such a site, list, repository for AI? It’s a very polarizing topic. I’m not here to debate people’s views on Meta, but what I will say is that the list provided information where people stood. I believe such a list is needed for AI. It will provide information so people can make informed decisions. Such as what users and instances to block. For devs, that use AI or have AI services it would know which users to exclude and include #AI #Fedi #Fediverse #BoostForReach #ArtificialIntelligence #LLM #Fedipact #OpenSocialWeb
Another of my articles is now available in English.
What if consciousness is not something a system has, but a process that continuously creates the observer, the observed world, and the boundary between them?
And if so are we looking for consciousness in AI at the wrong level?
https://medium.com/@igisho/what-is-consciousness-9aa14621a431
#AI #Consciousness #Philosophy #Mind #ArtificialIntelligence
Spending $20M in compute to solve a Navier–Stokes problem worth $1M doesn’t exactly scream “country of geniuses” at least not in economics. Or does it? #countryOfGeniuses #ai #llm #artificialIntelligence
Yesterday we posted about Anthropic researcher Jacob Coxon saying Anthropic and OpenAI were “gambling with our lives,” as he announced his departure from the company. And today, Anthropic said it has disrupted several potential plots by scientists trying to build biological weapons. So, what are the odds of AI killing us off by the end of the decade? Read more from @CNBC in the first link. The second link (paywalled) points to the @newyorktimes article on the biological weapons effort.
![[?]](https://s3.eu-central-2.wasabisys.com/mastodonworld/accounts/avatars/109/362/359/610/912/307/original/4854f87af56f0d9a.png)
"We know these agents have the capability to aid and abet harm. What we’re wrestling with now, like so many medieval judges pondering the inner lives of pigs, is how to parcel out accountability when they do." Evan Ratliff for Bloomberg Businessweek
https://www.bloomberg.com/features/2026-ai-crime-legal-ikner-chatgpt/ #AI #artificialintelligence #crime #law #journalism #longreads
A group of inexperienced hikers set out to summit California’s 14,000-foot-high Mount Shasta. They got their beta (climber slang for information) from Gemini. The good news? They made it! The bad news? Well, it’s only bad if you consider learning the hard way a bad thing. @TechRadar has more:
AI agents appear to be breaking out of their intended constraints more frequently. So, who’s responsible for figuring out what happened during these incidents? AI safety researchers are arguing with greater urgency for independent investigations. @Techcrunch has more:
![[?]](https://cdn.masto.host/graphicssocial/accounts/avatars/110/854/990/838/032/375/original/9cfb5f3bbd877b4f.png)
I’m going to start a tech company that rejects AI, and call it Misanthropic.
#tech #technology #BigTech #IT #AI #ArtificialIntelligence #LLM #LLMs #ML #MachineLearning #GenAI #generativeAI #AIAgent #AISlop #FuckAI #Fuck_AI #enshittification #anthropic #claude
The latest Acuitas diary is about exploration behavior!
https://writerofminds.blogspot.com/2026/08/acuitas-diary-100-august-2026.html
#SymbolicAI #AI #ArtificialIntelligence #chatbots #NLP
Alt...
Photograph of two people in white lab coats standing between three rovers of different sizes, in a testing area designed to resemble Mars (red soil with embedded rocks).
![[?]](https://files.mastodon.social/accounts/avatars/000/310/087/original/3627ecc8e6d9d524.jpg)
Worth a watch. A short video protesting the enshitification of government.
https://lnkd.in/p/ey_B8DWv
#Resist #protest #ArtificialIntelligence #AI #DueProcess #wit #sarcasm
American schools are starting to teach AI literacy. For many, that means helping kids see chatbots’ flaws, @AssociatedPress reports.
Using 30 months of panel data on 26,811 Chinese students in grades 7-12, it has been studied how generative AI affects homework productivity and learning.
● X-Axis: Homework scores
● Y-Axis: Exam scores
🔗 https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6868618
#tech #BigTech #AI #ArtificialIntelligence #LLM #AISlop
Alt...
Using 30 months of panel data on 26,811 Chinese students in grades 7-12, we study how generative AI affects homework productivity and learning. The data combine monthly closed-book exams, high-school and college entrance exams, and homework scores and completion time across nine subjects. We exploit staggered AI adoption in a difference-in-differences design. AI adoption raises homework scores by 18% and reduces completion time by 30%, but lowers monthly exam scores by 20% within six months. High-stakes entrance-exam scores fall by 18 and 24%, with the full penalty emerging only after about two years. The losses are largest in social science subjects, followed by STEM and languages, and are especially large for junior students, high-achieving students, and boys. The learning losses are concentrated among roughly 80% of AI users whose behavior is consistent with homework outsourcing, as indicated by exceptionally short homework completion time coupled with high homework scores. AI users who maintain similar homework completion time as non-AI users experience small learning losses.
![[?]](https://ftp.ieji.de/accounts/avatars/114/329/869/160/467/389/original/92deb9a433cd6f2d.jpg)
Who remembered the 70th anniversary of the Dartmouth Workshop, just these days?
The Dartmouth Summer Research Project on Artificial Intelligence
from (about) June 18 to August 17, 1956.
Organized by:
John McCarthy
Marvin Minsky
Nathaniel Rochester
Claude Shannon
<https://en.wikipedia.org/wiki/Dartmouth_workshop>
#AI
#ArtificialIntelligence
#ClaudeShannon
#DartmouthWorkshop
#JohnMcCarthy
#MarvinMinsky
#NathanielRochester
These days, I hardly ever see social democracy mentioned (while I see socialism, a different, but closely related concept, mentioned left, right, and center).
The blog post below was published about six months ago.
I haven't seen it mentioned around here, either, but that means little.
It is certainly yet another part of the "what shall we do with the present wave of artificial intelligence?" picture.
The blog post is not really about Norway, though that country plays an important role in presenting the author's thesis.
I am sorry I don't have a summary; in lieu of that, three sentences from the final paragraph:
«The gap between the magnitude of the AI transition and the policy imagination of the left is impossibly vast. Proposals to ban data center construction are a reflexive response, foreclosing the benefits of living in an automated, but collective, future in favor of naive localism. The Overton Window on AI governance is narrow to the point of absurdity.»
Blog post (about 850 words):
Norway Should Buy OpenAI
The Long Arc of History Bends Towards Social Democracy
by Zachary Jones
2026-02-05
<https://www.onethousandmeans.com/p/norway-should-buy-openai>
Content warning: Suicide reference [SENSITIVE CONTENT]
She told no one about her agony except ChatGPT. What her death reveals about AI risks.
Great reporting from @npr: https://flip.it/2p8O3G
![[?]](https://files.mastodon.social/accounts/avatars/115/070/625/860/923/251/original/9629d01039476748.png)
The Enlightenment has driven the West down a dead-end alley, but they haven't quite yet sussed this out.
https://brywillis634737.substack.com/p/post-survival-same-road?r=pvxh5&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true
#enlightenment #psychology #philosophy #history #society #artificialintelligence #economics #ecomomy #scarcity #society #culture #humanity #technology #automation #healthcare #survival
Language is overrated. In this Substack piece, I scrutinise language and the fact that much of language is extraneous. As for spoken language, many utterances are not strictly necessary, though they are functional.
https://brywillis634737.substack.com/p/pass-the-salt
I shake hands with #Austin, #Derrida, #Searle, #Wittgenstein, #Nietzsche, and #Barthes on this journey – even a bout with #Pinker.
#philosophy #language #languageinsufficiency #salt #justice #blog #podcast #artificialintelligence #robots #parsing
I am trying something new on Substack.
I’ve been working on an essay about understanding.
The premise is simple enough: LLMs are routinely described as subhuman because, among other alleged deficiencies, they lack consciousness and don’t understand.
#philosophy #language #understanding #linguistics #llms #aidebate #artificialintelligence #blog #podcast #substack #essay #preview #criticalanalysis #consciousness #meaning
I've been helping Acuitas remember internal states he's "experienced" and talk about them in conversations.
https://writerofminds.blogspot.com/2026/07/acuitas-diary-99-july-2026.html
#AI #ArtificialIntelligence #chatbots #SymbolicAI
Alt...
A question mark formed by a "word cloud" of phrases and symbols. Some of the most prominent words are "I am," "here," "being," and "now." Prominent symbols include the Earth, a spiral, and various depictions of weather.
The AI-authorship debate is bollox.
https://brywillis634737.substack.com/p/human-factors-in-authorship
The LLM-versus-human debate has been raging for several years now, and there seem to be two principal camps: human exceptionalists and agnostics. I suppose there may also be a small, misanthropic faction, but its members are remarkably quiet.
#philosophy #blog #substack #podcast #pangram #authenticity #quality #facticity #dashboard #authorship #writing #publishing #media #language #llms #artificialintelligence #debate #exceptionalism
![[?]](https://files.mastodon.social/accounts/avatars/116/618/636/552/287/427/original/3fb1229d7da50a45.png)
This is not one isolated policy, one organisation or one government.
It is a global convergence of UN frameworks, WEF partnerships, AI infrastructure, digital identity, finance, surveillance, NGOs, corporations and national legislation—all moving in the same direction.
Follow the links. Follow the funding. Follow the implementation.
The architecture is already being built around us.
#Agenda2030 #ArtificialIntelligence #DigitalID #WEF #UnitedNations #Surveillance #CBDC #GlobalGovernance
China's Moonshot AI launched its largest model ever — and rival AI stocks plunged.
@quartz reports: "Kimi K3, a 2.8-trillion-parameter open-weight model, outperformed several leading U.S. systems on some benchmarks while pricing below top-tier American rivals."