The use of Artificial Intelligence (AI) offers tremendous opportunities to improve business processes and customer experiences. In recent years, AI and Machine Learning (ML) have made significant advancements, with ChatGPT being a popular example. AI is in the spotlight, possessing the potential to bring about significant economic and societal changes across various sectors. However, amidst this potential, critical aspects must not be overlooked. AI also poses serious risks, particularly regarding ethical, legal, and operational facets. Both executives and developers, integrating AI into products and services, need to keep specific AI risks in mind. In Europe, a pivotal law is under discussion to address this. In April 2021, the European Commission presented the first draft for an EU legal framework for AI. The draft AI Regulation aims to establish EU-wide rules for the deployment of AI systems, which will have far-reaching implications for companies and software manufacturers in Europe.
These legal requirements might complicate and increase the cost of developing and operating AI-based systems. Early understanding and compliance with legal requirements and the resulting need for action offer a valuable competitive advantage.
Scope And Key Contents
The Commission draft proposal is designed as a horizontal EU legislative instrument applicable to all AI systems introduced or used in the Union. The central term of the AI Regulation is an AI system as defined in Article 3. According to the current Council Proposal [1], an AI system is defined as “[…] a system designed to operate with elements of autonomy based on data generated by machines and/or humans, employing machine learning and/or logic- and knowledge-based approaches to achieve a set of goals, producing results like content (generative AI systems), predictions, recommendations, or decisions that influence the environment with which the AI systems interact.” Applications falling within this definition are covered
by the regulation, necessitating a risk analysis, from
which legal obligations are derived. The AI Regulation specifies four risk classes: Unacceptable Risk (Article 5), High-Risk Systems (Articles 6 to 51), Low Risk (Article 52), and Minimal Risk (Article 69). “Unacceptable Risk”: AI systems presenting unacceptable risks are those violating EU values by endangering health, safety, or fundamental rights. Examples include cognitive behavioral manipulation and real-time remote biometric identification systems. Applications in this risk class will be prohibited, with violations leading to fines of up to €30,000,000 or up to 6% of global annual turnover. “High Risk” : High-risk AI systems can adversely affect people’s health, safety, or fundamental rights. They include safety components of products subject to EU product safety legislation and applications listed in Annex III. High-risk AI systems must meet extensive technical and organizational requirements, with providers needing to establish a quality assurance system encompassing comprehensive risk management. “Low Risk”: For certain low-risk AI systems, special transparency and information obligations are relevant, including labeling requirements and instructions. Users should be informed that they are interacting with a “No or Minimal Risk”: Such AI systems are exempt from the scope of application, subject only to general legal provisions, especially GDPR.
Implications For Software
Development And Need For Action
For system providers, the new legal framework presents both challenges and opportunities. Companies must ensure their products and services comply with new requirements and adjust their development processes, as the legal requirements affect the entire AI life cycle. A robust AI governance framework is crucial, offering a structured blueprint based on proven standards and best practices. This framework supports compliance with the AI Regulation by incorporating compliance checks and governance principles into the entire AI life cycle. While the final text of the law is still under discussion, system providers are advised to familiarize themselves with the potential requirements. Although a multiyear implementation period (planned between 24 and 36 months) is granted, early preparation is advisable due to the profound impacts and the importance of AI 1 Council Proposal: Proposal for a Regulation of the European Parliament and of the Council laying down harmonized rules on artificial intelligence (Artificial Intelligence Act) legislative acts, Council of the European Union, File 2021/0106(COD), Brussels, November 2022. 1. For implementation, it is recommended to follow best practices and standards such as MLOps or, specifically for data mining, CRISP-DM, ASUM-DM, or the Halerium Methodology. Particularly for the implementation of trustworthy AI systems, the ALTAI Assessment Tool provided by the High-Level Expert Group on Artificial Intelligence (AI HLEG) or the corresponding guideline from the Fraunhofer Institute offers good support.
About the Author
Armin D. Rheinbay Armin D. Rheinbay is responsible for risk management at Sopra Steria Next. With around thirty years of experience in the financial industry, he has worked as a consultant, entrepreneur, and project manager for the development of risk measurement methods and the implementation of company- wide risk management systems. Armin supports his clients at Sopra Steria in effectively and cost- efficiently mastering the risks associated with digital transformation while meeting regulatory requirements. He studied Corporate Finance, Modern Capital Market Theory, and Communication Sciences at the Universities of Göttingen and Hamburg, graduating with a degree in Economics.
A4Q Certified Selenium 4 Tester
The A4Q Testing Summit was recently held in Switzerland. This beautiful backdrop was the perfect platform for industry experts to come together and exchange knowledge and network. Engaging presentations were broadcast live on various social media channels with interactive sessions, allowing attendees to gain valuable insights. The A4Q Testing Summit was a resounding success and provided an unforgettable experience for all participants, whilst pursuing excellence in software testing. We can’t wait for the next Global Testing Summit! The ISTQB® General Assembly recently met in Tokyo, Japan. This exceptional assembly of dedicated volunteers shares a common enthusiasm for testing and commits their time to bolstering the continuous evolution of the ISTQB® Certified Tester program. Aheartfelt thank you to everyone for their active involvement and unwavering support! The prestigious ISTQB® Award celebrates those who have made exceptional contributions to software quality, through innovation, research, and promoting the software testing profession. The ISTQB® International Software Testing Excellence Award winner for 2023 is Mark Fewster! Mark Fewster’s nomination for this award recognizes his decadeslong dedication to advancing software testing. His involvement in ISTQB® committees, particularly for the Test Automation Engineer qualification, demonstrates his commitment to shaping industry Mark Fewster also created high-quality learning materials and introduced a licensing model that has enabled numerous global organizations to offer ISTQB® courses since 1999. These resources, accessible to over 40 licensees worldwide, have greatly influenced the careers and testing practices of countless software testers, solidifying Mark’s status as a luminary in the field. The Team behind SQmag Editor: Stephan Goericke Editor-in-Chief: Leonie Samuels Editorial and Design Team Lead Design: Claudine Readings Lead Copywriter: Liesl Hartje Send in your compelling content for our next publication. You are welcome to share your stories with us by email: contact@sq-mag.com