Jonathan Armstrong on The General Purpose AI Code of Practice [Podcast]

By Adam Turteltaub On July 10, 2025 the European Commission posted The General-Purpose AI Code of Practice.  Unlike the EU AI Act, this new Code of Practice is not compulsory, at least not yet. Still, it seems prudent to start understanding what it says and what expectations are being laid, as well as what the definition of general-purpose AI (GPAI) is.  To that end, we spoke with  London-based Jonathan Armstrong, Partner at Punter Southall. Jonathan explains that GPAI systems perform generally applicable functions such as image and speech recognition, audio and video generation, pattern recognition, question answering and translation.  It is similar to generative AI but is not the same. He then shares that the Code of Practice contains three sections:  transparency, copyright, and safety and security. Transparency is a hugely important issues for AI.  Organizations need to keep their technical documents related to their AI use current and address topics such as how the AI was designed, the technical means by which it performs functions and energy consumption. Copyright is a significant source of litigation at present.  Authors and other content creators see the use of their work by AI engines as a violation.  AI developers see the use of those works as furthering a greater good.  The Code of Practice sets out measures designed to help navigate these difficult waters. Safety & Security guidance is targeted predominantly at the most impactful GPAI operations.  The Code calls for extra efforts to examine cybersecurity and the impact of the technology.  This chapter of the document also includes 10 commitments for organizations to make. Listen in to the podcast and then spend some time reviewing The General-Purpose AI Code of Practice.   It’s worth seeing where regulations, and perhaps your AI efforts, are going.