IAPP AIGP, study guide domain ii
- Ley Muller
- Aug 21
- 19 min read
Domain ii: Understanding Laws, Standards, Frameworks Apply to AI Systems
Domain II.a How existing data privacy laws apply to AI
Relevant privacy laws concerning the use of data.
The Federal (US) Privacy Act of 1974 and the E-Government Act of 2002 require agencies to address the privacy implications of any system that collects identifiable information on the public
The Health Insurance Portability and Accounting Act (HIPAA)= health
Health Insurance Technology for Economic and Clinical Health Act of 2009 (HITECH), which increased penalties under HIPAA and provided greater access rights to individuals.
The Family Educational Rights and Privacy Act (FERPA) = student education
Protection of Pupil Rights Amendment of 1978 (PPRA) = student info
State privacy laws
CCPA/CPRA (CAlifornia!!),
Any AI tool used in customer-facing apps or analytics must honor these rights, or businesses risk lawsuits and fine:.
Informing users about what personal data is collected and why
Allowing users to opt out of data sharing or sales
Giving individuals the right to request data deletion
Maintaining clear privacy notices
Virginia Consumer Data Protection Act (VCDPA), CPA, CTDPA, Montana’s Consumer Data Privacy Act, Delaware Personal Data Privacy Act, Utah Consumer Privacy Act (UCPA), Oregon Consumer Privacy Act (OCPA), Iowa’s Consumer Data Protection Act (ICDPA), New Jersey Data Privacy Act (NJDPA), Indiana Consumer Data Protection Act, Tennessee Information Protection Act, Texas Data Privacy and Security Act (TDPSA)
Understanding key GDPR intersections
Data controller
Has primary obligations to data subjects — they must uphold subjects’ rights, such as allow Data Subject Access Requests (DSAR)
Must conduct Data Protection Impact Assessment (DPIA) / or Privacy Impact Assessment (PIA) if outside the GDPR
DPIA = risk assessment of a data processing activity, for the individuals whose data it is
PIA = slightly broader and more flexible. US, Canada, UK
Must ensure data processor compliance.
answer directly to supervisory authorities, and while audits extend to their processors, controllers face primary scrutiny.
Must establish a lawful basis for processing personal data (e.g., consent, contractual necessity) and communicate this to data subjects.
Data processor
Person separate from the controller, who processes personal data on its behalf.
must ensure that persons authorised to process the personal data have committed themselves to confidentiality (Article 28(3));
must maintain a record of all categories of processing activities
must implement appropriate technical and organisational measures
Understand automated decision making, DPIA, anonymization, and how they relate to AI systems
AI systems processes data - often personal data
“The data subject shall have the right NOT to be subject to a decision based solely on automated processing, including profiling, which produces legal effects concerning him or her or similarly significantly affects him or her.”
One effective control is to establish a human-in-the-loop procedure = to ensure human oversight and the ability to contest decisions.
UNLESS they consent
“the data controller shall implement suitable measures to safeguard the data subject’s rights and freedoms and legitimate interests, at least the right to obtain human intervention on the part of the controller, to express his or her point of view and to contest the decision.”
Creating a dataset to train an AI system may involve the use of PII, i.e. the proccessing of personal data. This can lead to “high risks” to people’s rights and freedoms. A DPIA is mandatory in this scenario. This is different than an AI Conformity assessment.
Anonymizing the training data can help to mitigate concerns by separating the information from the person.
Purpose limitation
= requires that personal data be collected for specified, explicit, and legitimate purposes and not further processed in a manner that is incompatible with those purposes.
Data minization
mandates that personal data collected should be adequate, relevant, and limited to what is necessary in relation to the purposes for which they are processed
GDPR data categories
Personal data
all information that can directly or indirectly identify an individual.
Special categories of personal data = must take extra precautions.
E.g. biometric data (facial images, voiceprints, iris scans, keystroke patterns) when used for the purpose of uniquely identifying individuals
Political views
Religion
Race
Sexual orientation
Union membership
Health -but health data CAN be processed by healht insurers and health crae professionals, with consent
Anonymous data
GDPR doesn’t apply
Pseudoanonymized data
GDPR still applies
People can be identified via a key
This is different under US law! Psuedoanonymization is enough
Understand the intersection between requirements for AI Conformity Assessments and DPIAs.
BOTH involve risk assessment and mitigation plans
“Before a high-risk AI system can be brought into the market, a Conformity Assessment (CA) should be made to ensure compliance with the AI Act
Regarding personal data
in the AI Act, the GDPR is explicitly mentioned when it comes to processing personal data], and when performing a DPIA is required.
This means there will be an overlap between these two assessments as a high-risk AI system would almost automatically include high risk processing under the GDPR”
But a high-risk AI system doesn’t have to involve personal data - in which case a DPIA Wouldn’t be involved
“A Conformity Assessment (CA) is focused on ensuring compliance with specific legal requirements, which are considered mitigation measures for high-risk systems.
The main goal of CA is to guarantee adherence to the mitigation measures or requirements mandated by the law.
An approved CA = required to enter the market.”
“A DPIA has a slightly different purpose. it serves as a tool for accountability by requiring controllers to assess and make decisions based on risks. It also mandates reporting on the decision-making process.
The primary objective == hold controllers accountable for their actions and ensure more effective protection of individuals’ rights.
The controller is ‘free’ to decide if and how it will mitigate risk.”
“Whenever a high-risk AI system involves the processing of personal data, a DPIA will almost certainly be required.
Both processes involve assessing risks related to specific systems and have distinct sets of requirements. To prevent redundant work or conflicting conclusions, it is probable that the CA can form the foundation for the DPIA of the controller.
If the provider also operates as a controller under the GDPR, then both the DPIA and CA will be carried out by the same entity, reinforcing each other.”
GDPR requirements for human supervision of algorithmic systems.
Users have the right to:
Know about the automated decision;
Understand the decision-making logic;
Challenge the decision and share their perspective; and
Request human intervention for decision review.
This is why data controllers must plan for human intervention, allowing individuals to review their situation, understand the decision, and contest it.
Understand an individual’s right to meaningful information about the logic of AI systems.
The existence of automated decision making, including profiling.
“Meaningful information about the logic involved.” Article 22 of the GDPR
should be understood as information around the algorithmic method used rather than an explanation about the rationale of an automated decision.
For example, if a loan application is refused, Article 22 may require the controller to provide information about the input data related to the individual and the general parameters set in the algorithm that enabled the automated decision.
But Article 22 would not require an explanation around the source code, or how and why that specific decision was made.”
“The significance and the envisaged consequences of such processing” for the individual.
"The privacy notice must be updated to explain the nature of automated processing, the logic involved, and the significance and consequences for the data subject.”
Domain II.b How other types of existing laws apply to AI
Understand the existing laws that interact with AI use -
IPR,
non-discrimination,
consumer protection,
product liability
Know the consumer protection laws that address unfair and deceptive practices.
In general: targeted advertising & recommendations are accepted, as long as they comply with transparency and consent requirements
regulated under consumer protection laws:
Deceptive claims
Biased financial decisions
Unauthorized data use
Federal Trade Commission (FTC) Act (US) (Wheeler-Lea Act of 1938)
EU Directive on unfair commercial practices from 2005
Children's Online Privacy Protection Act (COPPA) = governs the collection of information about minors
Gramm Leach Bliley Act (GLBA) = banks and financial institutions
Telemarketing Sales Rule (TSR), Telephone Consumer Protection Act of 1991, and the Do-Not-Call Registry
Junk Fax Protection Act of 2005 (JFPA)
Controlling the Assault of Non-Solicited Pornography and Marketing Act of 2003 (CAN-SPAM) and the Wireless Domain Registry
Telecommunications Act of 1996 and Customer Proprietary Network Information (CPNI)
Cable Communications Policy Act of 1984
Video Privacy Protection Act of 1998 (VPPA) and Video Privacy Protection Act Amendments of 2012
Driver's Privacy Protection Act (DPPA)
relevant non-discrimination laws (credit, employment, insurance, housing, etc.).
The Fair Credit Reporting Act (FCRA), which regulates the collection and use of credit information. regulates "consumer reporting agencies" and people who use the reports generated by consumer reporting agencies.
Crucially, a generative AI service potentially could meet the definition of "consumer reporting agency" if the service regularly produces reports about individuals' "character, general reputation, personal characteristics, or mode of living" and these reports are used for employment purposes.”
Confidentiality of Substance Use Disorder Patient Records Rule Prohibits patient information helping criminal charges
Fair and Accurate Credit Transactions Act of 2009 (FACTA)
contains protections against identity theft, “red flags” rules
Privacy Protection Act of 1980 (PPA)
The PPA requires law enforcement to obtain a subpoena in order to obtain First Amendment- protected materials
Title VII of the Civil Rights Act of 1964 prohibits employment discrimination on the basis of race, color, religion, sex, or national origin
Title I of the Americans With Disabilities Act (“ADA”) prohibits employment discrimination against “qualified” individuals with disabilities
Genetic Information Nondiscrimination Act of 2008
Illinois Artificial Intelligence Video Interview Act – Requires that any employer relying on AI technology to analyze a screening interview must provide information to candidates and obtain consent; must also report demographic data to the state to analyze bias
Maryland HB 1202 – Prohibits the use of facial recognition technology in the hiring process without consent of applicant
NYC Regulation – A bias audit must be conducted on any use of automated employment decision tools requires; notice must be provided to applicants and alternative selection process must be provided
The Wiretap Act
Know relevant product safety laws.
Consumer Product Safety Act (CPSA) in 1972 for the purposes of protecting consumers against the risk of injury due to consumer products, enabling consumers to evaluate product safety, establishing consistent safety standards, and promoting research into the causes and prevention of injuries and deaths associated with unsafe products.
EU - The General Product Safety Regulation requires that all consumer products on the EU markets are safe and it establishes specific obligations for businesses to ensure it. It applies to non-food products and to all sales channels.
relevant IP law.
Most important for copyright = originality
“U.S. Patent and Trademark Office (USPTO), U.S. Copyright Office, and courts have yet to fully establish clear guidelines concerning AI-created content or inventions. However, they generally recognize rights only for human authors and inventors.
European Patent Office (EPO) and European Union Intellectual Property Office (EUIPO) have similar stances, though discussions are ongoing about potential changes.”
Understand the basic requirements of the EU Digital Services Act (transparency of recommender systems).
The DSA imposes obligations on all information society services that offer an intermediary service to recipients who are located or established in the EU, regardless of whether that intermediary service provider is incorporated or located within the EU.
Transparency obligations: Advertising, user profiling, and recommender systems
Article 26 DSA, providers of online platforms must supply users with information relating to any online advertisements on its platform so that the recipients of the services can clearly identify that such information constitutes an advertisement.
Targeted ads based on profiling using special category data or personal data of minors = prohibited by providers of online platforms
Article 27DSA requires providers of online platforms that use recommendation systems to set out in their T&Cs the main parameters they use for such systems, including any available options for recipients to modify or influence them. Under Article 38, VLOPs and VLOSEs must provide atleast one option (not based on profiling) for users to modify the parameters used.
Understanding liability reform
= = the process of changing the legal rules and principles that govern the responsibility and accountability of parties who cause or contribute to damage or harm through AI systems.
Awareness of the reform of EU product liability law.
Needs to be updated because the current directive does not adequately cover digital services and connected products.
Article 4 of the proposed Directive brings software into the scope of EU product liability laws. Operating systems, firmware, computer programs and applications and AI systems are all expressly included (by Recital 12).
Article 7 extends liability to manufacturers of defective components, distributors, fulfilment service providers and online platforms.
Articles 8 and 9 provide a disclosure regime and set of rebuttable presumptions designed to assist claimants.
Understand the basics of the AI Product Liability Directive.
“Complements the Artificial Intelligence Act by introducing a new liability regime that ensures legal certainty, enhances consumer trust in AI, and assists consumers’ liability claims for damage caused by AI-enabled products and services.It applies to AI systems that are available on the EU market, or operating within the EU market.”
US federal level
EO14091= Further Advancing Racial Equity and Support for Underserved Communities Through the Federal Government (from 2023… rescinded in 2025)
required government institutions to assess the impact of AI-driven decisions on marginalized communities, ensuring these systems did not reinforce systemic bias.
The term “algorithmic discrimination” refers to instances when automated systems contribute to unjustified different treatment or impacts disfavoring people based on their actual or perceived race, color, ethnicity, sex (including based on pregnancy, childbirth, and related conditions; gender identity; intersex status; and sexual orientation), religion, age, national origin, limited English proficiency, disability, veteran status, genetic information, or any other classification protected by law.
Also revoked in 2025: Executive Order 14110: “The Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence - risk management and oversight
Domain II.c Existing and Emerging AI laws and standards
Understanding the requirements of the EU AI Act
Understand the classification framework of AI systems (prohibited, high-risk, limited risk, low risk).
Prohibited- considered a significant threat to fundamental rights, democratic processes, and societal values. Such systems are likely to compromise the integrity of critical infrastructures and engage in activities that could lead to serious incidents.
purposefully manipulative,
exploits vulnerabilities,
predatory targeting,
social scoring/profiling if it leads to negative outcomes,
predictive policing on the individual level,
untargeted facial image scraping meant to identify people,
emotion recognition in the workplace of workers (not customers),
biometrics if intended to deduce special categories of GDPR data (race, union membership, etc)
High Risk-
strict conformity assessments to ensure their accuracy, robustness, and cybersecurity, and their deployment is heavily regulated to mitigate potential risks associated with their use.
Mandated Human oversight here is because we want human judgement, expertise and understanding
when intended to be used as a safety components for medical devices, industrial machinery, toys, aircraft, cars, rail infrastructure, lifts, or appliances burning gaseous fuels,
Safety component of a medical device
The medical device itself
Biometrics (if not prohibited)
*out of scope: biometrics to enable cybersecurity
Access or admission to education/training - because of the impact on people’s livelihoods
Employment - recruitment or selection, or to make decisions that affect the terms of work relationships
Access to essential private and public services -
e.g. assessing eligibility of health care,
Evaluating credit, insurance
Emergency calls/services
Most law enforcement
Migration, asylum and border control management
Admin of justice and democracy (elections, courts, laws)
Limited Risk- considered less risky than their high-risk counterparts and thus face fewer regulatory constraints. However, while they do not require the same level of scrutiny, they must still adhere to specific transparency obligations to maintain accountability and trustworthiness in their deployment. This means that the developers and operators of these systems must be able to provide clear explanations of how the system works, what data it uses, and how it makes decisions.
No EU database registration, no CA
Chatbots
Minimal Risk: no mandatory obligations. The key aspect of this proposal is that it seeks to minimize regulatory burdens placed on such systems, thereby promoting innovation and development in areas where risks associated with the use of AI are deemed negligible or non-existent.
Examples : video games, spam filters, basic recommenders
requirements for high-risk systems and GPAI models exceeding systemic risk thresholds (10^25 FLOPs)
Demonstrating that the tech and its use does not pose a significant threat to health, safety and fundamental rights.
Conformity assessment
CE marking
Quality management system
Risk management
Data management o enable data governance
Post-market monitoring
Reporting system of serious incients
FHRIA
Technical documentation
Audits
Training and testing
logs
Human oversight
EU database registration
GPAI
All GPAI model providers must provide technical documentation (including training data summary), instructions for use, comply with the Copyright Directive, and publish a summary about the content used for training.
Free and open licence GPAI model providers only need to comply with copyright and publish the training data summary, unless they present a systemic risk.
GPA + systemic risk (open or closed) = providers must ALSO
conduct model evaluations,
adversarial testing,
track and report serious incident
ensure cybersecurity protections.
notification requirements (customers and national authorities).
“The AI Act requires developers/providers of high-risk AI to set up a reporting system for serious incidents as part of wider post-market monitoring.
A serious incident = as an incident or a malfunction that led to, might have led or might lead to serious damage to a person’s health or their death, serious damage to property or the environment, the disruption of critical infrastructure or the violation of fundamental rights under EU law.
developers/providers must notify market surveillance authorities within 2 days, 10 days (if a death), or 15 days, dependign on severity. Deployer can also notify.
Allowed to subject an initial incomplete report
Understand the enforcement framework and penalties for noncompliance.
Penalties for non-compliance follow a three-tiered system, with more severe violations of obligations and requirements carrying heftier penalties.
highest | Violations Prohibited systems | 35 million euro OR 7% turnover |
middle | Violations high-risk | 15 million euro OR 3% turnover |
lowest | Incorrect, incomplete, misleading info | 7.5 mill euro OR 1% turnover |
All actors can get a penalty for non-compliance = providers, deployers, importers, distributors, and notified bodies.
Understand procedures for testing innovative AI and exemptions for research.
An exception within the AI Act to process special categories of personal data to detect and correct bias within AI applies to providers of AI systems.
“the AI Act's exception applies to developers and entities outsourcing the development of AI systems, for non-private use. The exception does not seem to apply to organizations renting a fully developed AI system as a service, for example.”
Understand transparency requirements
PROVIDERS must register High-risk AI systems in an EU-wide public database
obligation to warn people that they are interacting with an AI system.
Understand other emerging global laws
Understand the key components of Canada’s AIDA, Artificial Intelligence and Data Act (C-27).
“AIDA provides a definition of “person” that includes trusts, partnerships, unincorporated associations and any other legal entity, and further clarifies when a such a “person” will be considered responsible for an AI system. A person becomes a “person responsible” for an AI system if they design, develop, make available for use, or manage the operation of an AI system in the course of international or interprovincial trade and commerce.”
Responsibilities include:
ensuring the anonymization of data
conducting assessments to determine whether an AI system is “high-impact,”
establishing measures related to risks
monitoring and keeping records on risk mitigation
requirements for organizations to publish a plain-language description of all high-impact AI systems on a public website.
If adopted, will replace PIPEDA
Minister of Innovation, Science and Industry must be notified about a high-impact AI system upon initial deployment
Understand the key components of U.S. state laws that govern the use of AI.
California, Connecticut, Vermont, Hawaii, Illinois, New York, Oklahoma, Rhode Island (lost steam), and Washington (lost steam)
“Provisions found in most of these bills require regular impact assessments of AI tools to ensure against discrimination; disclosure of such assessments to government agencies; internal policies, programs and safeguards to prevent foreseeable risks from AI; accommodating requests to opt-out of being subject to AI tools; disclosure of the AI's use to affected persons; and an explanation of how the AI tool uses personal information and how risks of discrimination are being minimized”
Automated employment decision tools- "predictive data analytics" used by employers to make employment decisions about hiring, firing, promotion and compensation.
Illinois, Massachusetts, New Jersey, New York, Vermont
“require employers to provide advance notice to and obtain consent from job applicants and employees who are subject to AEDTs, explain the qualifications and characteristics that AI will assess to candidates, and conduct and disclose regular impact assessments or bias audits of AI tools. Most of these bills, however, include carveouts for the use of AI when promoting diversity or affirmative action initiatives.”
AI Bill of Rights
“provide state residents the rights to know when they are interacting with AI, to know when their data is being used to inform AI, not to be discriminated against by the use of AI, to have agency over their personal data; to understand the outcomes of an AI system impacting them and to opt out of an AI system”
Oklahoma and New York
Working Group Bills
“creating government commissions, agencies or working groups to study the implementation of AI technologies and develop recommendations for future regulation”
Utah, Florida, Hawaii, Massachusetts
Understand the Cyberspace Administration of China’s draft regulations on generative AI.
apply to services offered to the public and NOT the use of genAI services by enterprises.”
During development, genAI providers must:
not generate illegal content such as false or harmful information;
prevent the generation of discriminatory content;
not use advantages in algorithms, data, or platforms where this leads to monopoly and unfair competitive behaviors;
not infringe on others’ portrait rights, reputation rights, honor rights, privacy rights and personal information rights; and
take effective measures based on service types to increase the transparency of generative AI services and the accuracy and reliability of generative AI content.
re: training data, generative AI service providers must:
use data and foundation models from legitimate sources;
not infringe others’ legally owned intellectual property;
obtain personal data with consent or under situations prescribed by the law or administrative measures; and
take effective measures to increase the quality of training data, their truthfulness, accuracy, objectivity and diversity.
When providing, generative AI service providers bear cybersecurity obligations as online information content producers and personal information protection obligations as personal information handlers and must:
enter into service agreements with registered generative AI service users which specify the rights and obligations of both parties;
guide users on the legal use of generative AI technology and take effective measures to prevent users from over-reliance on or “addiction to” the generated AI service;
not collect non-essential personal information, not illegally retain input information and usage records which can be used to identify a user and not illegally provide users’ input information and usage records to others;
receive and settle data subjects’ requests;
tag generated content such as photos and video as pursuant to the Administrative Provisions on Deep Synthesis of Internet-based Information Services (Deep Synthesis Provisions);
if illegal content is discovered, take measures to stop the generation and transmission of and delete illegal content, take rectification measures such as model improvement, and report to the relevant competent authorities;
where users are found to use generative AI services to conduct illegal activities, take measures to warn the user, or restrict, suspend or terminate the service, retain the records, and report to the relevant competent authorities; and
establish a mechanism for receiving and handling users’ complaints.
In relation to other legal obligations and enforcement supervision, generative AI service providers shall:
if the generative AI service comes with a public opinion attribute or social mobilization ability, carry out a safety assessment obligation and (within ten working days from the date of provision of services) go through record-filing formalities pursuant to the Administrative Provisions on Algorithm Recommendation for Internet Information Services (Algorithm Provisions); and
when the relevant competent authorities (e.g., the CAC) commence supervisory checks on the generative AI service, cooperate with them, explain the source, size and types of the training data, tagging rules and the mechanisms and principles of the algorithm and provide necessary technology and data, etc., for support and assistance.
Domain II.d Industry standards and tools that apply to AI
Understand the similarities and differences among the major risk management frameworks and standards
ISO 31000 Risk Management – Guidelines.
“A management system is the framework of policies, processes and procedures employed by an organization to ensure that it can fulfill the tasks required to achieve its purpose and objectives.”
Governance and culture; strategy and objective-setting; performance; information, communications and reporting; and the review and revision of practices to enhance the performance of the organization.
Emphasis on leadership endorsement and engagement, emphasis on organizational governance, emphasis on iterative nature of risk management (regularly updating processes and policies in response to new industry developments)
United States National Institute of Standards and Technology, AI Risk Management Framework (NIST AI RMF).
Voluntarily used to improve the ability to incorporate trustworthiness considerations into the design, development, use, and evaluation of AI products, services, and systems.
The framework breaks down the AI risk management process into four core functions: "govern," "map," "measure," and "manage”.
Govern = cross-fitting, enables the other three functions
establish a culture of risk management,
define roles and responsibilities,
develop and implement policies
Aim = comprehensive guidance on structures, systems, processes, teams
Map
Understand where AI is operating
Recognize context
= identify and understand AI risks within specific ocntexts & within AI lifecycle. This enables negative risk prevention.
Aim = introduce visibility, allow AI actors to see all parts of the process
Measure
= assess, analyze, track identified risks from MAP, to inform MANAGE
inclu de Testing, performance measurement, benchmarking,e tc
Manage
= prioritize risks and take action to mitigate them
Strategies to maximize AI benefits
Third party AI risks and benefits
seven “characteristics of trustworthy AI,” which include:
valid and reliable,
safe,
secure and resilient,
accountable and transparent,
explainable and interpretable,
privacy-enhanced, and
fair with harmful biases managed.
NIST ARIA = Assessing Risks and Impacts of AI program. Expands on AI RMF.
assess the societal risks and impacts of AI systems (i.e., what happens when people interact with AI regularly in realistic settings).
Classified as a testing, evaluation, validation and verification (TEVV) program, specifically within societal contexts after deployment.
Pilot test is focused on LLMs, and 3 levels of testing: model testing, red-teaming, and field testing.
European Union proposal for a regulation laying down harmonized rules on AI (EU AIA).
A proposed European law on artificial intelligence (AI) – the first comprehensive law on AI by a major regulator anywhere
The majority of obligations fall on providers (developers) of high-risk AI systems.
Users are natural or legal persons that deploy an AI system in a professional capacity, not affected end-users.
Council of Europe Human Rights, Democracy, and the Rule of Law Assurance Framework for AI Systems (HUDERIA).
Focus on risk and impact from the perspective of human rights, democracy,and rule of law
based on Council of Europe (CoE) standards
Four parts
Context-based risk analysis
Stakeholder engagement
Risk and impact assessment
Mitigation plan
IEEE 7000-21 Standard Model Process for Addressing Ethical Concerns during System Design
Braid in ethical values to systems engineering design/development
Make ethical values traceable. Operationalize ethics in system design.
Leadership + engineering + stakeholders
Relevant for all sizes and types of organizations, using their own life cycle models
ISO 22989
a common vocabulary for AI , to ensure consistency across other standards.
Not a certifiable standard.
Terminology baseline used by:
ISO/IEC 23894 — AI risk management
ISO/IEC 23053 — Framework for AI systems using machine learning
ISO/IEC 42001 — Artificial Intelligence Management System (AIMS)
ISO/IEC 24028 — Overview of trustworthiness in AI
ISO/IEC 5259 (series) — Data quality for analytics and ML
ISO/IEC/IEEE 29119-11 — Testing of AI-based systems
Does NOT say anything about third party procurement
Attribute | How it is addressed or framed in ISO 22989 |
Terms for performance consistency and dependable behavior across lifecycle phases | |
Concepts relating to harm, hazard, and safe operation of AI systems | |
Terminology linking security properties (confidentiality, integrity, availability) to AI contexts | |
Definitions around robustness to perturbations, uncertainty, and dataset shift | |
Concepts for recovery and continued operation under adverse conditions | |
Shared language for making AI system capabilities and limitations visible | |
Definitions for explainability/interpretability to support understanding of outputs | |
Roles and responsibilities, human oversight, assurance concepts | |
Terms for bias, fairness, and mitigation approaches | |
Concepts for data protection in AI lifecycles | |
Dataset, labeling, quality characteristics across training/validation/testing | |
Usability / human factors | Human-in/on/over-the-loop, human oversight terminology |
Lifecycle and change-related terms that support maintainable operation | |
Terminology for artifacts, provenance, and evidence across the lifecycle |
ISO 42001 AI Management System
Structured framework to orgs to develop or deloy AI system
Key components of an AI Management System include governance structures, risk management strategies, compliance protocols, and training programs to build competence among personnel involved in AI projects.
ISO 9001 quality management
ISO/IEC 42001 builds on this by providing a framework to manage the quality and consistency of AI systems.
Similarly, ISO 27001 information security
ISO/IEC 42001 these principles to the unique security risks and data protection challenges of AI.
ISO 13485 sets quality in medical devices
ISO/IEC 42001 supports this by ensuring AI components meet high standards of safety and effectiveness.
ISO/IEC Guide 51 Safety aspects – guidelines for their inclusion in standards.
“reducing risk that can arise in the use of products or systems, including use by vulnerable consumers. This Guide aims to reduce the risk arising from the design, production, distribution, use (including maintenance) and destruction or disposal of products or systems.”
ISO 42005 AI Impact Assessment
Singapore Model AI Governance Framework.
provides detailed and readily-implementable guidance to private sector organizations to address key ethical and governance issues when deploying AI solutions
Decisions made by AI should be: EXPLAINABLE, TRANSPARENT & FAIR
AI systems should be HUMAN-CENTRIC
recommends several measures to promote the responsible use of AI, such as determining the level of human involvement in decision-making, adapting governance structures, and establishing communications and collaboration among stakeholders
9 dimensionsof governance to foster a trusted AI ecosystem
Accountability
Data quality
Trusted dev and deployment (transparency)
Incident reporting
Testing and assurance (ideally third-party, external validation)
Security
Content provenance (transparency around origins)
Safety and alignment R&D
AI for public good
Comments