A Framework for Accountability: The Fundamental Rights Impact Assessment and the Future of AI Governance

The Fundamental Rights Impact Assessment (FRIA) under the EU AI Act is a critical component designed to ensure that the deployment of high-risk AI systems does not infringe upon the fundamental rights of individuals. This assessment is mandated for specific categories of deployers, including bodies governed by public law, private entities providing public services, and those using AI systems for creditworthiness evaluation, credit scoring, or risk assessment in life and health insurance.

Scope and Requirements

The FRIA must be conducted prior to the first use of a high-risk AI system. It involves a thorough analysis of the potential risks to fundamental rights, such as discrimination, privacy infringements, and restrictions on freedom of expression. The assessment includes several key elements:

• A description of the processes in which the AI system will be used.
• A description of the time frame and frequency at which the AI system will be used.
• Identification of the categories of natural persons and groups likely to be affected.
• An evaluation of the specific risks of harm to these individuals or groups.
• Implementation of human oversight measures and arrangements for mitigating these risks.
• Documentation and transparency measures to inform affected individuals and supervisory authorities.

Notification and Update

Deployers must notify the market surveillance authority about the results of the FRIA, using a template developed by the AI Office. If any elements of the assessment change during the use of the AI system, the deployer must update the information accordingly.

Integration with Existing Frameworks

The FRIA complements other impact assessments, such as the data protection impact assessment (DPIA) under the GDPR. While a DPIA focuses specifically on data privacy, the FRIA takes a broader view, assessing a wide range of fundamental rights including freedom of expression, access to justice, and the right to good administration.

Possible Challenges in Implementation

Despite its importance, the FRIA faces challenges in implementation. Deployers may struggle to fully assess the risks of high-risk AI systems, and effective methodologies for translating technical descriptions into concrete analyses of fundamental rights are still being developed. Additionally, there are concerns about the Act's shortcomings, such as the lack of explicit obligations to prevent negative impacts and exemptions for national security purposes, which could undermine the effectiveness of the FRIA in protecting human rights.

Conclusion

The FRIA is a crucial tool for ensuring that high-risk AI systems are deployed in a manner that respects and protects the fundamental rights of individuals. However, its successful implementation depends on addressing the current challenges and limitations inherent in the framework.

Resources-

• https://www.unesco.org/en/articles/ethical-impact-assessment-tool-recommendation-ethics-artificial-intelligence
• https://hbr.org/2020/10/a-practical-guide-to-building-ethical-ai
• https://oecd.ai/en/catalogue/tools/ai-ethics-self-assessment-questionnaire
• https://www.digitaldubai.ae/self-assessment
• https://www.digitalpolicy.gov.hk/en/our_work/data_governance/policies_standards/ethical_ai_framework/doc/Ethical_AI_Framework.pdf
• https://www.fujitsu.com/global/documents/about/research/technology/aiethics/fujitsu-AIethics-whitepaper_en.pdf

Co-Author: Shivang Mishra