International Standards Bodies Seek To Keep Pace With The AI Wave

Standards: the invisible infrastructure

getty

Standards can be called the “invisible infrastructure” – the guidelines and specifications that make our world, especially our technology world, tick. But can standards bodies keep up with the relentless week-by-week pace of change seen with artificial intelligence?

That’s the challenge for three of the world’s leading technical standards bodies, comprised of the International Electrotechnical Commission (IEC), the International Organization for Standardization (ISO), and the International Telecommunication Union (ITU).

I recently sat down with the secretary general of the IEC, Philippe Metzger, who discussed his organization’s efforts to develop standards for AI in all its various forms as quickly as the technology evolves. IEC, founded by Lord Kelvin in 1906, has been behind every standard over the past century, from fluorescent lamp safety to electrical currents for audio equipment.

The use of such standards are completely voluntary, Metzger emphasized. Adhering to standards in general, and especially AI standards these days, helps deliver competitive advantage, he said. “What we offer with our standards is a platform to compete. We contribute to global trade, and are recognized by the World Trade Organization. If you develop standards created by IEC and ISO, it increases your bankability, and trust from the market.”

The most prominent AI standard formulated and released to date is the ISO/IEC 42000 series, the most recent addition being ISO/IEC 42001:2023, which sets requirements and certifies compliance with an Artificial Intelligence Management System (AIMS).

IEC and its sister organization ISO are also formulating and releasing standards aimed at weeding out fake AI imagery – to develop and evangelize common standards to help users and companies distinguish erroneous AI-generated material from real content.

The JPEG Trust standard, announced last year by IEC and ISO, provides a framework for embedding metadata directly into JPEG files in the form of trust indicators. Two additions to the standard, now in progress, introduce a catalog of trust profile snippets and reporting templates, which can be used either as is or as starting points to establish profiles for use in specific workflows, use cases, and applications such as broadcasting, digital cameras, or AI-powered content generation services. Media asset watermarking is also provided.

In addition, EIC has been developing standards for introducing AI into industrial automation and cybersecurity and industrial processes, building on its ISO?EIC 27000 series of information security protocols, said Metzger. “That is being worked on by our technical experts and committees, to be reflected in a new version of the standard. We need to make sure that AI works within these systems that already exist, that they stay safe, and stay reliable.”

AI is changing fast, but Metzger points out that the standards being developed need to be “concrete” and well designed. “Our experts are working to achieve stability in technology. Importantly, the standards developed don’t specify particular technologies, but how to properly implement them. ”We identify the processes you need to apply internally to make sure everything is working properly," he said. “It doesn’t tell you what technology to use; it tells you these are the steps you need to take to make sure everything is efficient and ethical. It says these are the things you need to do. It doesn’t stipulate how you do them.”