Unveiled UK AI audit standard targets disruptive AI operators, aiming to bring order to the chaotic landscape
The British Standards Institution (BSI), the UK's national standards body, is set to revolutionise the artificial intelligence (AI) industry with the introduction of a new standard for companies that independently audit AI tools. This move aims to address concerns about the lack of standardisation and rigour in the AI assurance sector, as highlighted by the UK government.
The evolving landscape of AI standards and regulations is being shaped by various international standards such as ISO/IEC 42001, which focuses on artificial intelligence management systems. This standard emphasises the importance of robust AI risk management practices, including systematic testing protocols and auditable documentation to create evidence trails that demonstrate risk management and mitigation strategies.
ISO/IEC 42006, expected to be officially published in July 2025, sets out requirements for auditing bodies that certify against the ISO/IEC 42001 standard. This aims to ensure that certification bodies operate with the necessary competence and rigour, addressing concerns about independence and rigour in the AI assurance market.
The BSI's new standard will be the first international set of requirements to standardise how assurance firms check whether companies follow AI management standards. Philip Dawson, head of AI policy at AI assurance and insurance company Armilla AI, has stated that the BSI standard is a pivotal step forward for the AI assurance ecosystem.
The AI assurance market in the UK generates a gross value add of over £1bn, with boutique companies and larger contenders, including the Big Four accountancy firms, competing in the fledgling AI assurance market. However, many of these AI audit-selling groups also develop their own AI technologies, which raises concerns about their independence and rigor.
The BSI standard could potentially address these concerns, as it will clarify which companies are qualified to certify AI systems against ISO standards. This will help regulators, customers, and investors to differentiate between AI that has been assured by a certified assurance provider.
The demand for new AI assurance services is growing due to businesses' concerns about potential harms AI can cause and the need to comply with international regulations like the EU AI act. In the current landscape, hundreds of unchecked groups offer AI audits, claiming to assess AI models in various industries, such as self-driving cars and cancer detection.
Inioluwa Deborah Raji, a researcher at UC Berkeley who specialises in AI audits and evaluations, has pointed out concerns with proprietary AI audit standards. The BSI standard, with its focus on international requirements and rigorous accreditation, is expected to raise the bar for companies providing AI assessments and other certification bodies.
The launch of the new standard is scheduled for July 31, marking a significant step forward in the standardisation and regulation of AI assurance practices. This development is expected to foster trust and reliability in the AI industry, ensuring that AI systems are thoroughly understood, risk-tested, and appropriately controlled.
The British Standards Institution's new standard for companies that independently audit AI tools signifies a business revolution in the technology sector, particularly in the artificial-intelligence industry. This standard aims to address concerns about the lack of standardization and rigor in the AI assurance sector, and will establish international requirements to standardize how assurance firms check whether companies follow AI management standards.
With the BSI standard, as highlighted by researcher Inioluwa Deborah Raji from UC Berkeley, there is an expectation to raise the bar for companies providing AI assessments and other certification bodies, potentially addressing concerns about their independence and rigor in the AI technology business.
