This is not only unprecedented, but several factors elevate this continued standoff to an almost symbolic level in my view – the coincidence with intensifying wars is one, the confirmation of McLuhan’s thesis of the Medium being the Message. More specifically, he claimed that new media don’t just support war; they transform the structure of war itself. We see this play out in real-time.

So what happened within the past month? A lot, here the key facts:

Somewhat worryingly, but not unexpectedly in a year called the “year of agentic AI” by many, among them IBM’s Chief Architect Gabe Goodhart, real-life incidents are happening. CodeWallstill impressed with the extent – 100% of users of McKinsey’s AI platform Lilli, 728,000 files and 95 system prompts – and speed – 10 days from initial injection to public disclosure – of one of the broadest real-life examples of what can go wrong if technology is too many steps ahead of its governance.

From chatbots to assistants: governance is key for AI agents | World Economic Forum This WEF article looks at how the balance between autonomy and authority can be re-defined in contexts where agents take on increasingly operational tasks, increasing security requirements which might run counter to the seeming ease with which agents can be created, connected and operationalised.

The Guardian’s journalist Arwa Mahdawi summed up Karp’s statements towards CNBCin a way that is worth quoting directly: “Got that everyone? Disruption, disruption, disruption. And in case you didn’t catch it: disruption. Once you get beyond all the disruptive disruptions, it seems that what Karp is saying is that AI is eventually going to hurt the economic position of Democrats in general, and highly educated female voters in particular – and that will have knock-on effects politically. Meanwhile working-class male voters will emerge as the winners of our reshaped economy.” A few years ago, this would have been unthinkable, considered rude and blunt at least; now one has to assume the storyline Karp wanted to transmit has been carefully crafted.

On March 20th, the Trump administration published the National AI Legislative Framework. It consists of a four-page framework, the main part detailing six main objectives, such as protecting children online, enabling innovation and developing an AI-ready workforce. A critical side-topic is detailing out the initiative against “a patchwork of conflicting state laws (which) would undermine American innovation.” The Framework calls on Congress to establish a national standard that displaces state legislation targeting AI development, restricts AI activity that would be legal without AI, or penalises AI developers for third-party misuse of their systems. Major points stay unaddressed (classical points such as bias, privacy (beyond children), transparency, but also workforce displacement, procurement of AI systems) – and the initial driver behind the rise of federal AI regulation is left unaddressed.

Further reading tip: IAPP’s US-legislation tracker

Arthur Mensch lays out the status quo which, according to him, “satisfies no one”. He goes on to introduce a “revenue-based levy that would be applied to all commercial providers placing AI models on the market or putting them into service in Europe, reflecting their use of content publicly available online”. I am wondering: Is this (still) a way to go? Or was this a viable solution which wasn’t thought of enough when there was still practical leverage? Could this be a way forward? And how would this approach translate into the already complex digital regulation landscape? I would very much like Mensch’s statement “Europe does not need to choose between protecting its creators and competing in the AI race. It needs a framework that enables both” to prove true – and practically possible.

Meta and YouTube Found Negligent in Landmark Social Media Addiction Trial – The New York Times A Los Angeles jury found Meta and YouTube negligent on March 25, 2026, ruling that their addictive design features caused mental health harm to a young user — a landmark decision that could open social media companies to far more lawsuits. The plaintiff, a 20-year-old known as Kaley, alleged that using Instagram from age 9 and YouTube from age 6 led to addiction and serious mental health issues including depression, body dysmorphia, and suicidal thoughts. The jury awarded her $6 million in total damages. The case was strategically framed around the design of the platforms rather than the content on them — a legal approach that, if it holds on appeal, could reshape how thousands of similar pending lawsuits proceed.

The slightly absurd “AI adoption measured in tokens used” is examined in the Hard Fork Podcast. Some examples make me wonder: how desperate for measurable KPIs must one be to use something so obviously not indicative of value creation? Curious to see which alternative KPIs will be found and established in future to capture AI adoption contributing in a positive way!

In yet another interesting article, Microsoft’s Mustafa Suleyman argues for governance and against the anthropomorphisation of AI. AI is programmed to hijack human empathy — we must resist that

An insightful Politico article on where Europe is in terms of AI ambitions and which areas to address on the way there: Public sector AI: Shifting from ambition to readiness – POLITICO A detail that surprised me: more than half of public sector organisations are concerned about AI sovereignty – does this number (52% is what the data says) not seem a little low?