{"id":9201,"date":"2026-09-28T12:00:00","date_gmt":"2026-09-28T04:00:00","guid":{"rendered":"http:\/\/124.222.197.95\/?p=9201"},"modified":"2026-09-28T12:00:00","modified_gmt":"2026-09-28T04:00:00","slug":"iso-pas-8800-published","status":"publish","type":"post","link":"https:\/\/qinguard.cn\/en\/iso-pas-8800-published\/","title":{"rendered":"ISO\/PAS 8800 Is Published: Automotive AI Safety Now Has an International Standard, and the World's First Certification Has Been Issued"},"content":{"rendered":"<p><strong>Overview:<\/strong>In December 2024, the International Organization for Standardization <strong>(ISO) published ISO\/PAS 8800:2024, \"Road vehicles \u2014 Safety and artificial intelligence\"<\/strong>\u2014 the world's first international document dedicated to safety engineering for in-vehicle AI systems. In September 2026, the world's first ISO\/PAS 8800 certification was formally issued, marking this standard's transition from paper to the audit floor. This article covers its positioning, its scope of application, the safety evidence companies need to prepare, and the practical implications for Chinese companies.<\/p>\n<h2>1. What This Standard Is<\/h2>\n<ul>\n<li><strong>Standard number and title:<\/strong>ISO\/PAS 8800:2024, \"Road vehicles \u2014 Safety and artificial intelligence\";<\/li>\n<li><strong>Document type:<\/strong>PAS (Publicly Available Specification). Within the ISO system this is a document type with a shorter development cycle that can respond quickly to new technology needs, and which is later reviewed to decide whether it should be elevated to a full International Standard (IS);<\/li>\n<li><strong>Publication details:<\/strong>first edition, formally published on 13 December 2024, with approximately 172 pages of body text;<\/li>\n<li><strong>Responsible technical committees:<\/strong>ISO\/TC 22 (Road vehicles) \/ SC 32 (Electrical and electronic components), with ICS classifications 43.040.10 and 43.040.15.<\/li>\n<\/ul>\n<p>The number is easy to remember once broken down:<strong>ISO is the publishing body, PAS is the document type, 8800 is the number, and the subject is \"safety + artificial intelligence\".<\/strong>.<\/p>\n<h2>2. Why It Is Needed: Traditional Safety Standards Cannot Govern AI<\/h2>\n<p>The automotive industry's two existing safety heavyweights are <strong>ISO 26262 (functional safety)<\/strong>and <strong>ISO 21448 (safety of the intended functionality, i.e. SOTIF)<\/strong>. Both are powerful, yet both run into the same difficulty when faced with AI:<\/p>\n<ul>\n<li><strong>ISO 26262<\/strong> addresses \"deterministic systems\" \u2014 problems such as hardware failures and software logic errors that can be exhaustively enumerated and proven;<\/li>\n<li><strong>ISO 21448<\/strong> fills in the case where \"hazards can arise without a fault\", but it assumes that system behavior is a functional limitation that can be analyzed and described;<\/li>\n<li>Whereas <strong>AI systems are probabilistic<\/strong>: their behavior depends on the training data distribution, model architecture, training process and operating environment, and the same input may yield different results under different data versions.<strong>Reading the code and running test cases is not enough to demonstrate that it is safe.<\/strong>.<\/li>\n<\/ul>\n<p>This gives rise to three sources of AI-specific risk:<strong>data dependence (unsuitable or biased training data), black-box inexplicability (decisions that are hard to trace), and dynamic evolution (performance drift after deployment).<\/strong>ISO\/PAS 8800 establishes a structured framework for controlling these three categories of risk.<\/p>\n<h2>3. Scope: Whom It Covers and What It Covers<\/h2>\n<p>As defined in the standard itself:<\/p>\n<ul>\n<li><strong>What it covers:<\/strong>safety-related systems containing one or more E\/E (electrical\/electronic) systems that apply AI technology and are fitted to series-production road vehicles (excluding mopeds); E\/E systems unique to special-purpose vehicles (for example vehicles designed for drivers with disabilities) are out of scope;<\/li>\n<li><strong>Risks covered:<\/strong>unintended safety-related behavior at vehicle level arising from<strong>insufficient output, systematic errors or random hardware failures of in-vehicle AI elements<\/strong>; this also covers interaction with AI elements outside the vehicle \u2014 such external elements (for example off-board object detection or roadside monitoring) may affect vehicle safety directly or indirectly;<\/li>\n<li><strong>What it does not cover:<\/strong>development of the off-board AI elements themselves is outside the standard's scope, as are AI methods used as software tools;<\/li>\n<li><strong>Technical focus:<\/strong>the standard primarily targets<strong>machine learning (ML)<\/strong>methods and does not dwell on algorithm details (for example the architecture of a particular deep neural network); it focuses on how to demonstrate safety.<\/li>\n<\/ul>\n<p>The standard's objective is clear:<strong>to help companies build a persuasive Safety Assurance Claim demonstrating that the AI system does not present unreasonable risk.<\/strong><\/p>\n<h2>4. What Evidence Companies Must Produce<\/h2>\n<p>Translated into the questions an auditor would ask, the standard's requirements fall into roughly six categories:<\/p>\n<ol>\n<li><strong>Is the dataset suitable:<\/strong>do the training and validation sets represent real usage scenarios, and are the collection conditions and labeling process traceable?<\/li>\n<li><strong>Model performance:<\/strong>has the AI model's behavior been evaluated in normal, abnormal and edge-case scenarios?<\/li>\n<li><strong>Process management:<\/strong>are model training, testing and version management under controlled processes \u2014 can you explain which version changed, and why?<\/li>\n<li><strong>Unknown scenarios:<\/strong>could the system behave unexpectedly in situations never seen during training, and how are such cases identified and controlled?<\/li>\n<li><strong>After series production:<\/strong>are continuous monitoring and change management mechanisms in place once the vehicle is in service?<\/li>\n<li><strong>Evidence chain:<\/strong>do the test results and the safety claims form a complete, traceable body of evidence?<\/li>\n<\/ol>\n<p>In one sentence:<strong>ISO\/PAS 8800 does not regulate how a particular algorithm is written; it regulates the basis on which you claim your AI is safe \u2014 and whether that basis can be audited by a third party.<\/strong><\/p>\n<h2>5. The World's First Certification Has Been Issued<\/h2>\n<ul>\n<li><strong>Issuing and receiving parties:<\/strong>T\u00dcV Rheinland completed the world's first ISO\/PAS 8800 certification, with the R&amp;D Institute of<strong>Dongfeng Motor Corporation as the certificate holder<\/strong>;<\/li>\n<li><strong>Timeline:<\/strong>T\u00dcV Rheinland announced completion of the first certification in late August 2026; the certificate was formally presented on 15 September 2026 at the Dongfeng Motor Science and Technology Innovation Week event in Wuhan;<\/li>\n<li><strong>Scope of this certification (important):<\/strong>it covers the<strong>\"management process \/ management system\" for the safe development of in-vehicle AI systems<\/strong>\u2014 the control framework a company has built for AI-specific risks \u2014 and<strong>does not mean<\/strong>that all Dongfeng models or any particular automated driving function have obtained product safety certification.<\/li>\n<\/ul>\n<p>Why does this step matter so much? Because it answers the question the industry cares about most:<strong>can ISO\/PAS 8800 actually be audited, and how?<\/strong>The first certification sends three signals. First, the standard is implementable, and third-party bodies already have mature audit methods. Second, the audit examines data, models, processes and operational monitoring together \u2014 not merely code or accuracy metrics. Third, this kind of structured safety evidence will in future be directly linked to<strong>vehicle type approval and to discharging product liability<\/strong>\u2014 in other words, it is not just an internal technical document but an asset with legal and market significance.<\/p>\n<h2>6. China's Pace: From Benchmarking to Building Its Own AI Safety Standards<\/h2>\n<ul>\n<li><strong>Mandatory national standards already provide the foundation:<\/strong>three mandatory national standards \u2014 GB 44495\u20132024 \"Technical requirements for vehicle cybersecurity\", GB 44496\u20132024 \"General technical requirements for vehicle software updates\" and GB 44497\u20132024 \"Intelligent connected vehicles \u2014 Automatic driving data recording system\" \u2014 were published on 23 August 2024 and took effect on 1 January 2026, and were developed in full coordination with international regulations such as UN R155\/R156;<\/li>\n<li><strong>AI safety standards are accelerating:<\/strong>the Ministry of Industry and Information Technology's 2026 automotive standardization work priorities explicitly call for \"accelerating the development of foundational standards for intended functionality safety review and assessment, and for AI functional safety and safety of the intended functionality\", and set out a systematic plan for automotive AI standards covering AI risk assessment and governance, evaluation and testing of AI models for driving automation, automotive AI information security and data security, general development frameworks for end-to-end models, and intelligence grading;<\/li>\n<li><strong>The market-entry side has already tightened:<\/strong>the intelligent connected vehicle market access and road pilot program imposes multiple conditions on companies, including functional safety (ISO 26262 processes), safety of the intended functionality (SOTIF) and cybersecurity and software updates (the requirements corresponding to UN R155\/R156) \u2014 and these technical clauses overlap heavily with what ISO\/PAS 8800 focuses on.<\/li>\n<\/ul>\n<p>The conclusion is clear:<strong>international and domestic standards are tightening from both ends, and AI safety capability will become a precondition for intelligent driving products to be launched and exported.<\/strong><\/p>\n<h2>7. What Companies Should Do Now<\/h2>\n<ol>\n<li><strong>Do not build a parallel system:<\/strong>attach AI safety requirements to your existing functional safety \/ SOTIF organization and processes to form one traceable development chain, rather than standing up a separate system;<\/li>\n<li><strong>Start with data governance:<\/strong>dataset suitability, scenario coverage, and labeling and version records are the lowest-cost, highest-return area \u2014 and the one most likely to be challenged in an audit;<\/li>\n<li><strong>Build an AI safety file:<\/strong>thread datasets, models, test results and operational monitoring into an auditable safety argument, laying the groundwork for future type approval and liability defense;<\/li>\n<li><strong>Treat \"after launch\" as the starting point:<\/strong>establish model performance monitoring and change management to identify performance drift and new-scenario risks \u2014 precisely where AI differs most from traditional software;<\/li>\n<li><strong>Run a gap analysis:<\/strong>assess your current state against the standard's requirements, determine whether what you lack is process documentation, technical analysis or the evidence chain, then prioritize and advance in batches.<\/li>\n<\/ol>\n<p>For export-oriented companies there is one further reason: both the EU in automated driving approvals and the US in automated driving safety assessments are strengthening requirements for AI safety argumentation materials. Building this capability early is effectively securing a ticket to future markets.<\/p>\n<h2>Related Articles<\/h2>\n<ul>\n<li><a href=\"\/en\/iso-pas-8800-with-26262-21448-21434\/\">How to Use ISO\/PAS 8800: How It Works with ISO 26262, ISO 21448, and ISO\/SAE 21434<\/a><\/li>\n<li><a href=\"\/en\/iso-21434-overview\/\">What Is ISO\/SAE 21434: A Full Overview of the Automotive Cybersecurity Engineering Standard<\/a><\/li>\n<li><a href=\"\/en\/sotif-odd-mrm\/\">SOTIF, ODD and MRM: Core Concepts of Automated Driving Safety of the Intended Functionality<\/a><\/li>\n<li><a href=\"\/en\/autonomous-driving-triple-safety\/\">The \"Triple Triangle\" of Automated Driving Safety: Functional Safety, Safety of the Intended Functionality and Cybersecurity<\/a><\/li>\n<\/ul>\n<h2>Need AI Safety and Compliance Support?<\/h2>\n<p>If your organization is working on intelligent driving AI system safety, ISO\/PAS 8800, or ISO 26262 \/ ISO 21448 implementation, contact us for a free compliance diagnostic. QinGuard specializes in automotive safety and compliance consulting, providing end-to-end services from gap analysis and process development to safety argumentation and certification support.<\/p>","protected":false},"excerpt":{"rendered":"<p>\u3010\u5bfc\u8bfb\u30112024 \u5e74 12 \u6708\uff0c\u56fd\u9645\u6807\u51c6 [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"site-sidebar-layout":"default","site-content-layout":"","ast-site-content-layout":"default","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","ast-disable-related-posts":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"default","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"footnotes":""},"categories":[5],"tags":[12,13,59,61,60,35],"class_list":["post-9201","post","type-post","status-publish","format-standard","hentry","category-zidongjiashi","tag-iso-21448","tag-iso-26262","tag-iso-pas-8800","tag-tuv","tag-ai","tag-35"],"_links":{"self":[{"href":"https:\/\/qinguard.cn\/en\/wp-json\/wp\/v2\/posts\/9201","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/qinguard.cn\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/qinguard.cn\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/qinguard.cn\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/qinguard.cn\/en\/wp-json\/wp\/v2\/comments?post=9201"}],"version-history":[{"count":0,"href":"https:\/\/qinguard.cn\/en\/wp-json\/wp\/v2\/posts\/9201\/revisions"}],"wp:attachment":[{"href":"https:\/\/qinguard.cn\/en\/wp-json\/wp\/v2\/media?parent=9201"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/qinguard.cn\/en\/wp-json\/wp\/v2\/categories?post=9201"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/qinguard.cn\/en\/wp-json\/wp\/v2\/tags?post=9201"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}