🤖

ISO/IEC 23894

Information technology - Artificial Intelligence - Guidance on risk management

Organization: ISO/IEC Category: AI Risk Management
4
Related Articles
6
Articles with Obligations
6
Key Sections
9
Coverage Areas

Overview

Provides guidance for managing risks specific to AI systems including algorithmic bias, transparency, accountability, and safety. Critical for DSA platforms using AI-based recommender systems or content moderation (Articles 27, 35) to assess and mitigate AI-specific risks.

Applicability

AI system risk management for recommender systems and automated decision-making

Relevance to Digital Services Act (DSA)

Essential for DSA platforms using AI recommender systems (Article 27) and algorithmic risk assessments (Article 35)

Key Coverage Areas

1
AI-specific risk identification
2
Algorithmic bias and fairness
3
AI system transparency and explainability
4
AI safety and robustness
5
Data quality and governance for AI
6
Human oversight of AI systems
7
AI accountability and auditability
8
Recommender system risks
9
Automated content moderation risks

Standard Sections & Chapters

6

AI risk management framework

7

AI risk assessment

8

AI risk treatment

9

Monitoring and review of AI risks

Annex A

AI risk categories

Annex B

AI risk treatment options

Related Digital Services Act (DSA) Articles

Article 26: Advertising on online platforms

View Article →
Sections: 6, 7, Annex A

AI risk assessment methodology

Implementation Guidance:

Conduct AI-specific risk assessments for recommender systems

Article 27: Recommender system transparency

View Article →
Sections: 6, 8, 9

AI risk management and monitoring

Implementation Guidance:

Implement ongoing AI risk management for recommendation algorithms

Mapped Obligations:

  • Explain main parameters of recommender systems in terms and conditions using plain, intelligible language

Article 34: Risk assessment

View Article →
Sections: 6, 7

AI-specific risk assessment if applicable

Implementation Guidance:

Include AI/algorithmic risks in systemic risk assessment

Mapped Obligations:

  • Consider how recommender systems and algorithms influence risks

Article 35: Mitigation of risks

View Article →
Sections: 8

AI risk treatment measures

Implementation Guidance:

Apply AI-specific risk treatments for algorithmic systems

Mapped Obligations:

  • Test and adapt algorithmic systems including recommender systems

Quick Information

Organization
ISO/IEC
Category
AI Risk Management
Certification
Not available

🤝 Still Feeling Overwhelmed?

EU cybersecurity laws can be complex. Our free tools and guides work great for most people, but if you're dealing with something particularly challenging or have tight deadlines, we're here to help.