{"version":"https://jsonfeed.org/version/1.1","title":"AI Standards Blog","home_page_url":"https://aistandards.blog","feed_url":"https://aistandards.blog/feed.json","description":"On how intelligent systems should behave, where they fail, and who answers for them.","items":[{"id":"https://aistandards.blog/shadow-ai-discovery-and-control","url":"https://aistandards.blog/shadow-ai-discovery-and-control","title":"Shadow AI: discovery and control","summary":"If staff use a model on your time, devices, or data, you are already the deployer. Find consumer apps, off-label enterprise features, embedded suite assistants, and tool-using agents. Bin each finding: retire, replace, or record. A policy without a hunt is not control.","content_text":"If staff use a model on your time, devices, or data, you are already the deployer. Find consumer apps, off-label enterprise features, embedded suite assistants, and tool-using agents. Bin each finding: retire, replace, or record. A policy without a hunt is not control.","date_published":"2026-10-26T08:20:06.827+00:00","date_modified":"2026-10-26T16:31:49.164689+00:00","authors":[{"name":"Anwesh Rath"}],"image":"https://aistandards.blog/shadow-ai-discovery-and-control/opengraph-image"},{"id":"https://aistandards.blog/what-a-board-should-ask-about-ai","url":"https://aistandards.blog/what-a-board-should-ask-about-ai","title":"What a board should ask about AI","summary":"Ask for an inventory with owners, roles, tools, and the next legal date. Ask which agents can bind the company without a human. Refuse “we are fine until 2027.” Live duties already include Article 5, Article 50, and ordinary liability. Residual risk belongs in a sentence, in the minutes.","content_text":"Ask for an inventory with owners, roles, tools, and the next legal date. Ask which agents can bind the company without a human. Refuse “we are fine until 2027.” Live duties already include Article 5, Article 50, and ordinary liability. Residual risk belongs in a sentence, in the minutes.","date_published":"2026-10-26T08:20:02.61+00:00","date_modified":"2026-10-26T14:21:34.431376+00:00","authors":[{"name":"Joe Quenneville"}],"image":"https://aistandards.blog/what-a-board-should-ask-about-ai/opengraph-image"},{"id":"https://aistandards.blog/prompt-injection-ciso-playbook","url":"https://aistandards.blog/prompt-injection-ciso-playbook","title":"Prompt injection: a CISO playbook","summary":"Prompt injection turns untrusted text into instructions. On an agent with tools, that is a use of your privileges. OWASP lists no complete fix. Cut tools, run as the user, mediate outside the model, tag retrieved content, put a human on write/send/pay, and keep a replay log.","content_text":"Prompt injection turns untrusted text into instructions. On an agent with tools, that is a use of your privileges. OWASP lists no complete fix. Cut tools, run as the user, mediate outside the model, tag retrieved content, put a human on write/send/pay, and keep a replay log.","date_published":"2026-10-26T08:19:58.479+00:00","date_modified":"2026-10-26T14:21:46.006132+00:00","authors":[{"name":"Anwesh Rath"}],"image":"https://aistandards.blog/prompt-injection-ciso-playbook/opengraph-image"},{"id":"https://aistandards.blog/when-to-run-models-on-hardware-you-own","url":"https://aistandards.blog/when-to-run-models-on-hardware-you-own","title":"When to run models on hardware you own","summary":"Own the box when a leak, a silent model swap, or a vendor outage would break a core or regulated process. A vendor API is enough when the use is reversible and the data is already allowed to leave. Sovereignty is artefact, upgrade clock, and perimeter — not a flag on a VPC.","content_text":"Own the box when a leak, a silent model swap, or a vendor outage would break a core or regulated process. A vendor API is enough when the use is reversible and the data is already allowed to leave. Sovereignty is artefact, upgrade clock, and perimeter — not a flag on a VPC.","date_published":"2026-10-26T08:19:54.551+00:00","date_modified":"2026-10-26T14:21:42.477316+00:00","authors":[{"name":"Anwesh Rath"}],"image":"https://aistandards.blog/when-to-run-models-on-hardware-you-own/opengraph-image"},{"id":"https://aistandards.blog/model-provenance-and-weight-hashing","url":"https://aistandards.blog/model-provenance-and-weight-hashing","title":"Model provenance and weight hashing","summary":"A weight hash identifies the bytes you loaded. A model card does not. OWASP LLM03 says published models lack strong provenance and tells you to use signing and file hashes. If you cannot name the artefact, you cannot replay a decision or answer discovery.","content_text":"A weight hash identifies the bytes you loaded. A model card does not. OWASP LLM03 says published models lack strong provenance and tells you to use signing and file hashes. If you cannot name the artefact, you cannot replay a decision or answer discovery.","date_published":"2026-10-26T08:19:50.188+00:00","date_modified":"2026-10-26T14:21:38.643091+00:00","authors":[{"name":"Anwesh Rath"}],"image":"https://aistandards.blog/model-provenance-and-weight-hashing/opengraph-image"},{"id":"https://aistandards.blog/iso-42001-explained-for-the-board","url":"https://aistandards.blog/iso-42001-explained-for-the-board","title":"ISO/IEC 42001 explained for the board","summary":"ISO/IEC 42001:2023 is a voluntary AI management-system standard: policy, roles, risk, controls, audit, improvement. It does not certify that a model is safe and it does not replace the EU AI Act. A certificate is only as strong as its scope. The board still owns appetite, ownership, and evidence.","content_text":"ISO/IEC 42001:2023 is a voluntary AI management-system standard: policy, roles, risk, controls, audit, improvement. It does not certify that a model is safe and it does not replace the EU AI Act. A certificate is only as strong as its scope. The board still owns appetite, ownership, and evidence.","date_published":"2026-10-26T08:19:46.115+00:00","date_modified":"2026-10-26T14:21:30.680489+00:00","authors":[{"name":"Joe Quenneville"}],"image":"https://aistandards.blog/iso-42001-explained-for-the-board/opengraph-image"},{"id":"https://aistandards.blog/who-is-liable-when-an-ai-agent-acts","url":"https://aistandards.blog/who-is-liable-when-an-ai-agent-acts","title":"Who is liable when an AI agent acts?","summary":"The company that put the agent into service under its authority is the first person the law names. Vendors share exposure as providers, and software can be a product after 8 December 2026. Prompt injection does not move the loss. The model is not the defendant.","content_text":"The company that put the agent into service under its authority is the first person the law names. Vendors share exposure as providers, and software can be a product after 8 December 2026. Prompt injection does not move the loss. The model is not the defendant.","date_published":"2026-10-26T08:19:41.726+00:00","date_modified":"2026-10-26T14:21:26.076447+00:00","authors":[{"name":"Joe Quenneville"}],"image":"https://aistandards.blog/who-is-liable-when-an-ai-agent-acts/opengraph-image"},{"id":"https://aistandards.blog/eu-ai-act-2026-2027-by-role","url":"https://aistandards.blog/eu-ai-act-2026-2027-by-role","title":"What the EU AI Act requires in 2026–2027","summary":"The EU AI Act already binds most companies on bans, literacy, general-purpose models, penalties, and transparency. The July 2026 Digital Omnibus postponed high-risk system duties to 2 December 2027 for Annex III uses and 2 August 2028 for Annex I products. Role — provider or deployer — decides what you owe.","content_text":"The EU AI Act already binds most companies on bans, literacy, general-purpose models, penalties, and transparency. The July 2026 Digital Omnibus postponed high-risk system duties to 2 December 2027 for Annex III uses and 2 August 2028 for Annex I products. Role — provider or deployer — decides what you owe.","date_published":"2026-10-26T08:19:36.059+00:00","date_modified":"2026-10-26T14:21:21.722468+00:00","authors":[{"name":"Joe Quenneville"}],"image":"https://aistandards.blog/eu-ai-act-2026-2027-by-role/opengraph-image"},{"id":"https://aistandards.blog/state-of-ai/the-ai-agent-problem-who-is-actually-responsible","url":"https://aistandards.blog/state-of-ai/the-ai-agent-problem-who-is-actually-responsible","title":"The AI Agent Problem Nobody Wants to Discuss: Who Is Actually Responsible?","summary":"When an AI system begins acting on behalf of an organization, intelligence becomes only one part of the problem. The other part is accountability.","content_text":"When an AI system begins acting on behalf of an organization, intelligence becomes only one part of the problem. The other part is accountability.","date_published":"2026-10-02T06:05:00+00:00","date_modified":"2026-10-26T14:22:15.471026+00:00","authors":[{"name":"Anwesh Rath"}],"image":"https://kqfgumdgnjukrkgwhcos.supabase.co/storage/v1/object/public/media/uploads/2026-10-26/composite-f379cc79fa17c68a.jpg"},{"id":"https://aistandards.blog/state-of-ai/ai-is-becoming-invisible","url":"https://aistandards.blog/state-of-ai/ai-is-becoming-invisible","title":"AI Is Becoming Invisible. That Might Be Its Most Important Evolution.","summary":"The most consequential AI in your life may eventually be the AI you never consciously interact with. Which is precisely why its invisibility should not mean its decisions are beyond scrutiny.","content_text":"The most consequential AI in your life may eventually be the AI you never consciously interact with. Which is precisely why its invisibility should not mean its decisions are beyond scrutiny.","date_published":"2026-09-18T03:45:00+00:00","date_modified":"2026-10-26T14:22:13.111538+00:00","authors":[{"name":"Anwesh Rath"}],"image":"https://kqfgumdgnjukrkgwhcos.supabase.co/storage/v1/object/public/media/uploads/2026-10-26/composite-b1d91802cbf7fc05.jpg"},{"id":"https://aistandards.blog/state-of-ai/we-are-teaching-machines-to-think-are-we-forgetting-how","url":"https://aistandards.blog/state-of-ai/we-are-teaching-machines-to-think-are-we-forgetting-how","title":"We Are Teaching Machines to Think. Are We Forgetting How?","summary":"We are investing enormous resources into making artificial intelligence more capable. Perhaps we should devote a little more attention to keeping human intelligence intellectually active.","content_text":"We are investing enormous resources into making artificial intelligence more capable. Perhaps we should devote a little more attention to keeping human intelligence intellectually active.","date_published":"2026-09-04T05:20:00+00:00","date_modified":"2026-10-26T14:22:18.577488+00:00","authors":[{"name":"Anwesh Rath"}],"image":"https://kqfgumdgnjukrkgwhcos.supabase.co/storage/v1/object/public/media/uploads/2026-10-26/composite-fdb7891a0c11eeaa.jpg"},{"id":"https://aistandards.blog/state-of-ai/ai-agents-google-io-2026-are-we-ready","url":"https://aistandards.blog/state-of-ai/ai-agents-google-io-2026-are-we-ready","title":"AI Is Starting to Act, Not Just Answer. Are We Ready for That?","summary":"Google I/O 2026 marked a shift from AI that answers questions to AI that can perform tasks. That sounds convenient, but before handing an agent your email, money, or business systems, there are a few things you should understand.","content_text":"Google I/O 2026 marked a shift from AI that answers questions to AI that can perform tasks. That sounds convenient, but before handing an agent your email, money, or business systems, there are a few things you should understand.","date_published":"2026-05-25T04:10:00+00:00","date_modified":"2026-10-26T15:52:45.426295+00:00","authors":[{"name":"Anwesh Rath"}],"image":"https://kqfgumdgnjukrkgwhcos.supabase.co/storage/v1/object/public/media/uploads/2026-10-26/gio-google-io-host-d805c3b8b01a5c6d.jpg"},{"id":"https://aistandards.blog/building-sovereign-ai-the-distance-between-a-vision-and-a-working-product","url":"https://aistandards.blog/building-sovereign-ai-the-distance-between-a-vision-and-a-working-product","title":"Building Sovereign AI: The Distance Between a Vision and a Working Product","summary":"Sovereignty is not branding. It is a set of operational decisions and tradeoffs.","content_text":"Sovereignty is not branding. It is a set of operational decisions and tradeoffs.","date_published":"2026-03-04T09:30:00+00:00","date_modified":"2026-10-26T14:21:58.620283+00:00","authors":[{"name":"Anwesh Rath"}],"image":"https://kqfgumdgnjukrkgwhcos.supabase.co/storage/v1/object/public/media/ora/2026-10-26/5e8a89d7-1c09-4dcd-a590-bf78a993d77e-anwe-11ac4b443c8f0414-building-sovereign-ai-the-distance-betwe.png"},{"id":"https://aistandards.blog/the-architecture-of-an-ai-that-knows-what-it-doesn-t-know","url":"https://aistandards.blog/the-architecture-of-an-ai-that-knows-what-it-doesn-t-know","title":"The Architecture of an AI That Knows What It Doesn't Know","summary":"Good AI architecture makes uncertainty visible and enforceable.","content_text":"Good AI architecture makes uncertainty visible and enforceable.","date_published":"2026-03-01T09:30:00+00:00","date_modified":"2026-10-26T14:21:55.547817+00:00","authors":[{"name":"Anwesh Rath"}],"image":"https://kqfgumdgnjukrkgwhcos.supabase.co/storage/v1/object/public/media/ora/2026-10-26/a8ba49fe-333a-4752-b530-bdfa39888a30-anwe-141f404dd91d58c3-the-architecture-of-an-ai-that-knows-wha.png"},{"id":"https://aistandards.blog/a-blockchain-can-preserve-a-wrong-answer-forever","url":"https://aistandards.blog/a-blockchain-can-preserve-a-wrong-answer-forever","title":"A Blockchain Can Preserve a Wrong Answer Forever","summary":"Immutability establishes a record. It does not establish truth.","content_text":"Immutability establishes a record. It does not establish truth.","date_published":"2026-02-27T09:30:00+00:00","date_modified":"2026-10-26T14:22:07.498694+00:00","authors":[{"name":"Anwesh Rath"}],"image":"https://kqfgumdgnjukrkgwhcos.supabase.co/storage/v1/object/public/media/ora/2026-10-26/11e8a851-344c-42f0-a1fc-dd089602bd97-anwe-ea288a4c33cbd2cb-a-blockchain-can-preserve-a-wrong-answer.png"},{"id":"https://aistandards.blog/the-problem-with-ai-that-always-says-yes","url":"https://aistandards.blog/the-problem-with-ai-that-always-says-yes","title":"The Problem With AI That Always Says Yes","summary":"An assistant that refuses to acknowledge limits is not helpful. It is dangerous to depend on.","content_text":"An assistant that refuses to acknowledge limits is not helpful. It is dangerous to depend on.","date_published":"2026-02-05T09:30:00+00:00","date_modified":"2026-10-26T14:21:52.006541+00:00","authors":[{"name":"Anwesh Rath"}],"image":"https://kqfgumdgnjukrkgwhcos.supabase.co/storage/v1/object/public/media/ora/2026-10-26/92a7463e-5f37-46df-a171-811b5c6662c7-anwe-cff7ab045f5507b9-the-problem-with-ai-that-always-says-yes.png"},{"id":"https://aistandards.blog/why-i-started-building-ai-systems-instead-of-just-using-them","url":"https://aistandards.blog/why-i-started-building-ai-systems-instead-of-just-using-them","title":"Why I Started Building AI Systems Instead of Just Using Them","summary":"Using AI is easy. Making it dependable in a real workflow is another discipline.","content_text":"Using AI is easy. Making it dependable in a real workflow is another discipline.","date_published":"2026-02-02T09:30:00+00:00","date_modified":"2026-10-26T14:22:01.657303+00:00","authors":[{"name":"Anwesh Rath"}],"image":"https://kqfgumdgnjukrkgwhcos.supabase.co/storage/v1/object/public/media/ora/2026-10-26/61df5b4f-0bde-47e8-91a3-2bea9c013f1a-anwe-d12778dcef6ea65f-why-i-started-building-ai-systems-instea.png"},{"id":"https://aistandards.blog/why-ai-implementations-fail","url":"https://aistandards.blog/why-ai-implementations-fail","title":"Why Most AI Implementations Fail (And How to Fix It)","summary":"After 17 years in enterprise systems, I've seen the same patterns destroy AI projects over and over. Here's what actually works.","content_text":"After 17 years in enterprise systems, I've seen the same patterns destroy AI projects over and over. Here's what actually works.","date_published":"2026-01-28T10:00:00+00:00","date_modified":"2026-10-26T14:22:04.237898+00:00","authors":[{"name":"Anwesh Rath"}],"image":"https://aistandards.blog/why-ai-implementations-fail/opengraph-image"}]}