News article • 4 mins read
UKAS is pleased to support the AI Standards Stack podcast, a series from the AI Quality Infrastructure Consortium (AIQI) exploring how standards, assurance and governance can help build trust and confidence in Artificial Intelligence (AI).
As AI systems become more widely adopted across sectors, organisations are increasingly looking for credible ways to demonstrate that they are being developed, deployed and governed responsibly. AI assurance has an important role to play in this landscape, helping organisations, regulators, customers, investors and the public understand whether appropriate controls, standards and oversight are in place. The AI Standards Stack podcast is hosted by Professor Michael Mainelli, UKAS Non-Executive Director, and Adam Leon Smith, Chair of the AIQI Consortium. Across the series, they have been joined by a range of expert guests to explore the practical, technical and ethical questions shaping the future of AI assurance.
UKAS has supported several episodes in the series, featuring guests including Christine Chow (London School of Economics), Piercosma Bisconti Lucidi (DEXAI), Nicholas Beale (City of London Corporation), Patrick Sullivan (A-LIGN Certification), James Gealy (Safer AI) and Michael Thieme (Accenture). Together, these episodes provide a broad view of the evolving AI assurance landscape, from responsible AI governance and investor expectations through to certification, AI safety standards and the practical use of international standards.
In one episode, Christine Chow considers responsible AI governance through an investor lens, highlighting why robust data governance is a foundation for effective AI governance. This perspective is particularly relevant as organisations face increasing scrutiny over how AI systems are managed, how risks are controlled, and how governance arrangements can provide confidence to stakeholders beyond the organisation itself. The series also explores the ethical and social dimensions of AI. Piercosma Bisconti Lucidi discusses generative AI, ethics and the future of social interaction, including the implications of AI systems entering human social spaces and the risks that may arise when AI systems interact with one another. Nicholas Beale brings a mathematically grounded perspective to responsible AI, considering the risks of over-reliance on single AI systems, the limitations of human-in-the-loop approaches, and the importance of appropriate guardrails.
Other episodes focus more directly on the developing assurance infrastructure around AI. Patrick Sullivan provides a certifier-led view of AI assurance and ISO/IEC 42001, exploring what third-party certifiers do, how AI management system certification can support objective assurance, and why organisations are increasingly looking to certification as part of their approach to AI governance and risk management. In an episode with Michael Thieme, the series also looks at how standards can be made practical for professionals working with AI, discussing the need for “consumable” standards that can realistically be applied in the real world, particularly as organisations respond to rapid developments in generative AI. His episode considers issues such as AI verification, validation, performance measurement, reproducibility, terminology and the changing role of human oversight as agentic AI systems evolve.
A recent episode with James Gealy has explored AI safety standards for frontier models, including risk modelling work on how advanced models could amplify cyber threats, contribute to benchmark saturation and create real-world harms. The discussion also considers developing standards work on large language model benchmarking, red teaming and the safety of AI models, systems and applications.
Taken together, the series provides a useful and accessible overview of many of the issues now shaping AI assurance. It considers not only how AI systems can be designed and governed responsibly, but also how assurance mechanisms such as standards, certification, conformity assessment and accreditation may contribute to wider trust and confidence.
UKAS’s support for the series reflects its wider interest in the development of robust, credible and internationally relevant approaches to AI assurance. Accreditation has a key role to play in helping ensure that conformity assessment activities are delivered competently, consistently and impartially. As AI assurance continues to mature, the quality infrastructure will be important in supporting confidence in both AI systems and the organisations that assess them.
For organisations involved in evaluating, assessing or governing AI systems, the AI Standards Stack podcast provides useful insight into the evolving role of standards, accreditation and quality infrastructure in supporting credible AI assurance.
Listen to the series here: https://www.aiqi.org/newsroom/category/Podcast