Blog Post

Written by
Holistic AI
Category
Structured content
Date
June 6th, 2022
EU AI Act Primer
Flag of the European Union

Purpose

  • The European Commission introduced the Act to establish a regulatory framework within the Union for the development of AI.
  • The purpose of the Act is to create an ‘ecosystem of trust’ that manages AI risk and prioritizes human rights in the development and deployment of AI.

 

Duties on enterprises

  • All AI deployment is subject to transparency obligations to declare to users that they are interacting with an AI and allow users to opt-out.
  • High-risk AI deployment is subject to risk-management obligations to:
  • Establish risk-management systems through the AI system’s lifecycle;
  • Validate, test, and train data to ensure their quality;
  • Provide detailed documentation for third-party assessment;
  • Declare accuracy, robustness and security strategies for the mitigation of systemic failure;
  • Ensure appropriate human oversight of AI systems.

 

What are high-risk AI systems?

  • Systems are high-risk when they threaten the enjoyment of individuals’ basic rights and interests.
  • The AI Act only lists a number of use-cases as high-risk:
    • Biometric identification
    • Management of critical infrastructure
    • Educational assessment and access
    • Employment decisions
    • Access to essential services (private and public)
    • Law enforcement
    • Migration and border control
  • The Act empowers the Commission to extend this list. Additions are likely to focus on AI affecting the civil rights of individuals.

Who’s in scope?

  • Any company using AI systems that are on the market or produce outputs used on the market in the EU.
  • This scope includes US-based enterprises with operations affecting EU citizens.

 

What’s the risk?

  • Noncompliant enterprises face administrative fines from 4 to 6% of global revenue, and from EUR 20 to 30 million.

 

Strategic evaluation

  • Access to the European market is critical for enterprises in the digital economy.
  • As with GDPR, European standards are likely to be adopted across a broad range of jurisdictions, which means that compliance with the EU AI Act is likely to be of wider importance.

 

AI Act compliance with Holistic AI

  • HolisticAI’s research team has published and consulted widely on the EU AI Act and is firmly embedded in the European regulatory ecosystem.
  • Although there are other services offering governance support for EU compliance, HolisticAIhas the unique capacity to evaluate companies’ technical and governance systems for compliance with the incoming legislation.
  • HolisticAI’s is at the forefront of risk-mitigation innovation, allowing companies to manage risk sufficiently for compliance whilst optimizing their technical capacities.
  • HolisticAI’s subscription package for on-going monitoring and support offers companies a cost-effective solution to life-cycle management of AI.
Interested in our company?