On Wednesday 13 March 2024, the European Parliament endorsed the world’s first major legal framework on Artificial Intelligence. The AI Act aims to provide developers and deployers with clear requirements and obligations regarding the specific use of Artificial Intelligence by classifying systems in accordance with a risk-based approach. The Act is the first step in the delivery of a wider package of policy measures from the EU and sovereign governments to support the development of trustworthy Artificial Intelligence.
The Impact of AI Across Industries
Artificial Intelligence is profoundly changing the way organisations operate. The use of advanced analytics and generative algorithms to mimic and derive human behaviours enables organisations to scale cognitive abilities, delivering dramatic benefits to many industry sectors, particularly those dependent on customer-centric experience including financial services, insurance, pharmaceuticals, retail, and transport as well as in the public sector services such as healthcare, education, and environmental protection. As the following examples demonstrate:
• Artificial Intelligence is enabling the ‘hyper-personalisation’ of retail banking services to improve timely decisioning of loans and mortgages whilst also providing ethical support for customers requiring financial aid.
• Artificial Intelligence is enabling more precise coverage and pricing in insurance as well as streamlining underwriting processes, reducing claims fraud and developing new solutions for previously uninsurable risks.
• Artificial Intelligence is leading to a fairer distribution of healthcare services and the use of advanced analytics and algorithms is accelerating the modelling and treatment of complex conditions.
The Challenges of AI
However, Artificial Intelligence also introduces new challenges and threats - there are a growing number of cases that give reasons for concern. This not only threatens trust in AI-enabled services but also heightens public fears about personal data protection/privacy, intellectual property, and the human race itself through the weaponisation of technology.
• In 2016, Microsoft introduced the Tay chatbot on Twitter; the chatbot started with a friendly first tweet announcing ‘Hello World’ but rapidly developed an aggressive personality with racist, fascist and sexist views.
• In 2021, a Tesla Model S missed a curve in the road leading the vehicle to crash into a tree killing two people; Preliminary investigations suggested the driver’s seat was empty during the crash leading to speculation that the AI-based Full Self Driving (FSD) system was engaged during the crash.
• As recently as February 2024, Air Canada was forced to honour a refund policy its customer services chatbot had inaccurately cited after the airline had argued that it shouldn’t be liable for anything the chatbot says.
The Need for Ethical AI
When focused on customer needs, it is indisputable that Artificial Intelligence can deliver significant business benefits and service the public good. However, without appropriate controls, Artificial Intelligence can rapidly undermine brand reputation, as well as amplify concerns about job displacement, ethical bias, privacy, security, and a lack of human-like empathy. Ethical Artificial Intelligence must be fair, robust, transparent, accountable, and contestable and that requires a degree of public control over how and when we use artificial intelligence.
Quaylogic's Role in AI Governance
The EU AI Act goes some way to identify unacceptable risks but consistent and robust regulation still lags behind the advancement of technology. Quaylogic has been working with a number of clients to mitigate risk associated with Artificial Intelligence and the associated data that service advanced analytics and generative algorithms.
This approach identifies the domain classifications and risk categories to enable the organisation to define appropriate risk appetite, metrics, tolerances, and control principles. In many cases, this represents an extension to existing technology and data risk frameworks that additionally focus on mitigating the risk of fraudulent misuse and/or unintentional consequences of rogue technology as well as promoting transparency, governance, and ethical use of algorithms and associated data.
Quaylogic also continues to contribute to public sector initiatives that provide guidance and further support the risk-based approach. For example, The UK Government Department for Science, Innovation & Technology has published a number of research and consultation papers including initial guidance for Regulators entitled ‘Implementing the UK’s AI Regulatory Principles’ . This document sets out five principles for regulators to interpret and apply within their remit, and highlights some of the controls that should be considered in any risk framework covering safety, security, robustness and transparency as well as fairness, accountability, governance, contestability and redress.
Also a complementary guidance is provided by publications such as the ‘Algorithmic Transparency Standard’ and ‘Data Ethics Framework’ issued jointly by the UK Government Central Digital & Data Office (CDDO) and Centre for Data Ethics & Innovation (CDEI).
Unlocking the Full Potential of AI with Quaylogic
Quaylogic takes a leading role in developing frameworks and standards, as well as promoting awareness of some of the challenges and threats posed by Artificial Intelligence. This includes participation in industry bodies that are collaborating on innovative practices and ethical considerations for regulatory compliance, mitigating risk, building trust and exploiting competitive advantage through the development of standards and best practices as well as tools for maturity assessment, training and development.
Are you ready to unlock the full potential of AI in your data? 🚀
Contact Quaylogic today for expert support and guidance on adopting a risk based approach to Artificial Intelligence and Data. Our team is here to help you navigate the evolving landscape and drive transformative outcomes for your business.