Map: Identifying Risks, Benefits and Affected People

Lesson concept diagram
Who is affected: Identify risks and benefits for people

The Map phase of the NIST AI Risk Management Framework involves identifying risks, benefits, and affected people associated with an AI system. This step is essential for understanding the potential impact of AI deployment and ensuring that governance efforts are aligned with organisational goals and regulatory expectations. The focus here is on gathering factual data and stakeholder perspectives to inform decision-making in later phases of the framework.

Risk Identification

Organisations must identify potential risks tied to their AI systems through a structured approach. These risks can be technical, operational, legal or ethical. For example, an AI system used for credit scoring may pose a risk of bias if it disproportionately affects certain demographic groups. In such cases, the risk is tied to fairness and compliance with anti-discrimination laws. Another risk might be data privacy breaches, particularly if the AI system processes personal data without proper safeguards. The EU AI Act Article 5 prohibition on certain AI practices applies from 2 February 2025, which makes it important to identify any prohibited uses early in the process.

  • Unfair treatment of individuals or groups
  • Data protection or privacy violations
  • Unreliable or inaccurate outputs
  • Loss of human oversight or control
  • Unintended consequences due to system misuse

Organisations should also consider risks related to transparency and explainability. The EU AI Act Article 50 transparency duties apply from 2 August 2026, requiring AI systems to provide information about their use. For generative AI systems placed on the market before this date, machine-readable marking must be implemented by 2 December 2026. This highlights the importance of documenting AI systems early to meet these obligations.

Benefit Analysis

Identifying benefits helps organisations understand the value that AI systems bring. These benefits must be weighed against identified risks. For example, an AI-based clinical decision support tool may reduce diagnostic errors and improve patient outcomes. In such cases, benefits include clinical effectiveness, cost efficiency, and improved care delivery. However, these must be balanced against potential risks such as over-reliance on AI or misinterpretation of outputs by clinical staff.

Other benefits may include operational efficiency, automation of repetitive tasks, or enhanced decision-making through data analysis. The EU AI Act’s Article 4 AI literacy duties apply from 2 February 2025, which underscores the importance of training staff to understand AI systems. This investment in AI literacy can lead to better adoption, reduced misuse, and stronger compliance outcomes.

Organisations must also consider broader societal or business benefits. For example, AI systems that support sustainability goals or reduce environmental impact may align with corporate social responsibility strategies. These benefits should be quantified or at least clearly articulated to support decision-making and reporting.

Affected People

Understanding who is affected by an AI system is central to responsible AI governance. Affected people may include end users, data subjects, employees, or members of the public. For example, an AI tool used in recruitment may affect job applicants, or an AI-based surveillance system may impact citizens’ privacy rights. The EU AI Act requires that AI systems meet transparency and accountability standards, which means that affected individuals must be informed of AI usage where applicable.

Organisations should identify these groups through stakeholder mapping or consultation processes. For instance, a bank deploying AI for fraud detection must consider the impact on customers who may be incorrectly flagged or denied services. Similarly, a healthcare AI tool must consider the impact on patients, clinical staff, and clinical governance structures.

Other affected groups may include regulators, auditors, or third-party vendors. The EU AI Act’s GPAI model obligations apply from August 2025, and Commission enforcement powers begin from 2 August 2026. These developments mean that organisations must be prepared to demonstrate compliance to national supervisory authorities. The Digital Omnibus on AI, Regulation (EU) 2026/1744, defers high-risk AI obligations to 2 December 2027, which allows time for organisations to prepare but also highlights the importance of early planning.

ISO/IEC 42001:2023 provides guidance on establishing an AI management system. The first UKAS-accredited certification body was BSI on 15 January 2026, indicating that formal certification is becoming a standard practice. This underscores the importance of documenting AI systems, including who is affected, what risks exist, and what benefits they bring. The framework must be applied consistently across all AI systems, regardless of size or complexity, to ensure alignment with both organisational and regulatory expectations.

Custom Enterprise Workshop • The Tesseract Academy

In-House Corporate AI Strategy & Governance Clinics

Bespoke executive clinics led by Dr. Stylianos Kampakis for leadership teams, boards, and division directors.