If anything, it would be an understatement to suggest that the ascendancy of AI in broadcast and media applications has been rapid since SMPTE established a dedicated AI Taskforce in 2020. In fact, its trajectory has been nothing short of phenomenal, taking even Frederick Walls, AI Taskforce co-chair and AMD fellow, by surprise.
“Absolutely I’ve been surprised by the pace of development, especially when I think back to 2020 when not that many people were talking about AI,” he recalls. “From conversations with various people, I would get the impression that some companies had a couple of staff looking at AI in the corner and seeing what it might be able to do.”
But if it all felt rather tentative back then, AI in broadcast has moved decisively from theory to practice during the last two years. Hence why SMPTE, in conjunction with the European Broadcasting Union (EBU) and Entertainment Technology Center (ETC), has now updated the Engineering Report that it first issued in 2023, acknowledging the fact that “the AI in media landscape is evolving at an unprecedented pace, and the SMPTE AI Taskforce remains committed to keep up with this transformation,” notes SMPTE standards director Thomas Bause Mason.
In particular, the new report covers multiple recent developments around standards and technical frameworks. “One of the recommendations we made in this most recent revision of the report was to sort of standardise on things like Model Context Protocol (MCP) and A2A,” explains Walls. “Then, within about two weeks, Google and a number of big AI companies announced that they were going to standardise MCP.”
Likened by its developers to a “USB-C port for AI applications”, MCP is an open-source standard for connecting AI applications—such as Claude or ChatGPT—to external data sources, tools and workflows. Regarded as a complementary rather than competing development, the Agent-to-Agent (A2A) protocol enables AI agents “to communicate with each other, securely exchange information, and coordinate actions on top of various enterprise platforms or applications,” say its developers.
The report also addresses the latest developments around ISO/IEC 42001, the first international certification standard for Artificial Intelligence Management Systems (AIMS). Published in late 2023, the standard provides guidance to organisations that provide, produce or use AI systems, and can help them achieve compliance with legislation such as the EU AI Act and individual national legislation like Italy’s AI Law, the notably comprehensive piece of legislation which entered into force in October 2025.
In short, the foundational building blocks that will allow AI to develop in a standardised and interoperable fashion are beginning to fall in place. “If you want interoperability and for everyone to be able to use something without fear of intellectual property issues and so on, it’s best to go through a standardisation project with due process,” says Walls. “Making sure that everyone’s on the same page and doing things the same way is very important; without that, you can end up with security issues and all kinds of unexpected problems.”
The fact that progress is being made in multiple areas of standardisation is just as well, given that “industry-wide we are now seeing a huge level of interest in AI and, increasingly, a breakneck pace of change. From our perspective as SMPTE, we’re doing all we can to keep our members and others who are reading the report up to date with what is a very rapid pace of evolution,” says Walls.
Revolutionising the newsroom
Application-wise, Alexandre Rouxel – who is senior project manager, data and AI at the EBU – agrees that it is tasks such as localisation services and camera automation where deployment is most advanced. “For real-time transcription and captioning, it’s becoming the norm,” he says. “We are reaching a plateau [for these applications] where there will still be small improvements, but in many ways it has reached maturity.”
Rouxel is also keeping a keen eye on the development of AI tools in the newsroom for production and fact-checking. “We are now at the age of edge computing, which involves AI running locally for production, [in other words] on your devices and a lot of things that are more or less linked to the race of the LLMs [Large Language Models] and everything related to production.”
Also focusing on the newsroom, Walls anticipates further innovations around automation and quality control. “You already see vendors coming out with live production tools that are AI-enabled, so for example, you might have the news anchor saying something and then the software underneath sort of reacting to that. We see a lot of opportunities for improving the quality and efficiency on the broadcasting side of things, and think that people are recognising opportunities and starting to bring products to market in those areas. It’s a value proposition that basically aims to help people do what they do, but do it better.”
Meanwhile, the SMPTE AI Taskforce is an ongoing entity, with members usually meeting once a month to “talk a bit about the various projects going on in SMPTE, and the other areas where SMPTE could make a difference, for example, are there other opportunities for standards or recommended practices? So there are plenty of strands to keep on track and ensure that we can provide our members with quality information that helps them navigate this world of AI.”
Inevitably, this will also involve further revisions to the Engineering Report. “After this latest version came out, I asked the Taskforce whether we should already be thinking about the next one, and the answer was unequivocally ‘yes’,” says Walls. “So it could be that there will be another edition in two years’ time, which underlines how intense the pace of change has become.”
To download SMPTE Technology Reports, including the latest revision to the AI and Media Engineering Report, please visit https://www.smpte.org/technology-reports-downloads.
- This article originally appeared in the March issue of TVBEurope, available to download here
