For over three years, A IEEE Standards Association The Working Group is refining a draft standard for the purchase of Artificial Intelligence and Automated Decision System, IEEE 3119-2025Its purpose is to help identify and manage the purchase teams risk In high -risk domains. Such systems are used by government institutions included in education, health, employment and many others. Public area Area. Last year, the Working Group participated with an European Union Union Agency to evaluate the components of draft standard and collect information about users’ needs and their views at the value of the standard.
At that time, the standard included fiveProcedures to help users develop, reduce and monitor users usually identify, reduce and monitor the damage connected. high risk AI System.
Those processes had the definition of problem, vendor evaluation, solution evaluation, contract negotiations and contract monitoring.
The reaction of the European Union agency inspired the work group to reconsider the processes and the sequence of several activities. The final draft now involves an additional process: solicitation preparation, which comes immediately after the problem definition process. The working group believes that the additional procedure addresses the experience of challenges organizations with the preparation of AI-specific solicitation, such as adding transparent and strong data requirements and requires to include questions about the maturity of the seller AI regime.
The European Union agency also emphasized that it is necessary to include the preparation of solidity, which gives the procurement teams additional opportunities to customize their solidity with questions about technical requirements and responsible AI system options. Leaving space for adjustment is particularly relevant when AI is being acquired Emerging And Change the environment,
Gissle Waters
IEEE 3119 ecosystems in standards
There are currently many internationally sanctioned standards AI management, AI ethicsAnd normal Software acquisitionFrom IEEE Standards Association and International Organization for Standardization Goal AI design, use and life-cycle management.
So far, no international level has been accepted, the general consent-based standard focuses on the purchase of AI devices and provides operationsGuidance for shopping responsibly High -risk AI systems It serves public interest.
IEEE 3119 are standard addresses which intervals. Unlike AI Management Standards ISO 42001 and generic AI oversite and Risk regime, IEEE provides a new standard A of IEEE Risk-based, operationsApproach To help customize government agencies traditional Procurement practice.
Governments have an important role in the responsible deployment of AI. However, market dynamics and uneven AI expertise between industry and government may be obstacles to discourage success.
One of the main goals of the standard is to better inform procurement leaders before making high -risk AI purchases of what they are purchasing. IEEE 3119 defines high-risk AI systems that are sufficient factors in making or making Resulting decision It can have significant effectsOn people, groups, or society. The definition is similar to the one used in 2034 in Colorado Ai actThe first US state-level laws address a widely high-risk system.
Standard procedures, however, complement the ISO 42001 in many ways. The relationship between the two is painted below.
International standards, often characteristic Soft lawAI is used to shape development and Encourage international cooperation About its rule.
AI, or legally difficult laws for binding rules and obligations, are a task in progress worldwide. In the United States, A Scatter State law controls various aspects of AI, and national AI is an approach to regulation DisgracefulWith various federal agencies applying their own guidelines.
Europe has led by passing the European Union Ai actWhich started controlling the AI ​​system based on its level of risk when implemented last year.
But the world is lacking Regulatory hard law With an international scope.
The IEEE 3119-2025 standard aligns with current difficult laws. Due to focus on procurement, the standard supports the high -risk provisions mentioned in the standard Chapter III of EU AI Act And AI Act of ColoradoStandard is also proposed to propose Texas HB 1709 Legislation, which will make reporting compulsory on the use of AI systems by some commercial institutions and state agencies.
Because Most AI systems used in public sector Instead of in-house are purchased, IEEE 3119 applies to commercial AI products and services that are not required Adequate amendment Or adaptation.
Standard targeted audience
The standard means:
- Middle-level purchases with professionals and members of the interdisciplinary team with a medium level of AI rule and AI system knowledge.
- Public and private-field purchases are professionals who serve as coordinators or buyers, or have equivalent roles within their institutions.
- Non-processing managers and supervisors who are either responsible for the purchase or oversee employees who provide procurement functions.
- Professional who are employed by controlling public education, utilities, transportation, and other publicly funded services, who either work or buy on purchase and are interested in adopting purchasing processes for AI tools.
- AI sellers are trying to understand new transparency and disclosure requirements for their high -risk commercial products and solutions.
Working program
IEEE Standards Association Participated with AI Procurement Lab To offer IEEE responsible AI Purchase Training ProgramThe curriculum includes how to apply the main processes of the standard and optimize current practices for the purchase of high -risk AIs.
The standard includes more than 26 tools and rubricics in six processes, and the training program explains how to use many of these devices. For example, training includes instructions to conduct risk-copy analysis, apply seller assessment scoring guide to analysis AI seller claimsAnd create an AI procurement “Risk Register” that is associated with usage-caste risks and their potential mitigation. The training session is now available for purchase.
This is still an early day for AI integration. Decision manufacturers have not yet experienced much experience in buying AIs for high -risk domains and reducing those risks. The IEEE 3119-2025 standard aims to support agencies and strengthen their AI risk muscle muscles.
From your site articles
Related articles around web
