At Rise of AI 2026 in Berlin, Chris Boos argues that Europe has the lifestyle, security and knowledge base to lead in AI but is failing to act, citing 27 unimplemented national AI strategies against clear US and Chinese approaches. He frames sovereignty as the ability to control dependencies, not autarky, since no economy can be fully self-sufficient, and argues Europe's real asset is knowledge, not energy or raw materials, which it has never turned into a shared platform. Boos walks through a data-information-knowledge-intelligence-wisdom hierarchy, arguing that data is never "true" but a propagation of interactions through a system, and that Europe over-focuses on data transport instead of adding semantic, causal meaning that lets machines and humans work together on complex networks. He describes large language models as an interface technology comparable to the web browser, letting non-experts operate complex systems, and argues agentic AI's main risk to work is not job loss but the loss of "training ground" through which experts traditionally build skill. He closes urging curiosity over security, community over self, facts over feelings, and treating people with intrinsic respect regardless of role.
Key takeaways
- Chris Boos argues Europe has 27 national AI strategies but has executed none of them, unlike the US and China.
- He defines sovereignty as controlling dependencies through counter-dependencies, not as unreachable autarky.
- He argues data is never objectively true because it propagates through a system at different speeds; it only shows a propagation state.
- He frames large language models as an interface technology, comparable to the web browser, that lets non-experts operate complex systems.
- He argues the real risk of agentic AI to work is the loss of the training ground through which experts traditionally develop skill, not job loss itself.
Detailed notes
Europe's stalled progress despite its advantages
Chris Boos describes moving from Silicon Valley back to Europe for lifestyle and security reasons, then criticises the audience for not using Europe's advantages. He contrasts AI's investment hype, exemplified by Microsoft, Nvidia and OpenAI investing circularly in each other, with a genuine underlying shift, and points to Apple's much higher revenue per employee than Volkswagen as evidence that companies applying technology well are pulling ahead of those that do not. He recalls helping write a German AI strategy that shrank from a 3-page to a 13-page to a 30-page paper with no content left, contrasting this with implemented strategies in the US and China and "27 AI strategies in Europe" with no execution.
Sovereignty as controlling dependencies, not autarky
Boos argues access to technology is also a way of exercising power, citing China's rare-earth exports, Russian energy and US model access as potential leverage points, and that every dependency needs a counter-dependency so a threat can be answered with "don't even try." He says Europe conflates sovereignty with autarky, which is unreachable in a complex, interdependent world, and quotes Henry Kissinger's line that the US has interests, not friends or allies, as advice Europe should heed. He argues Europe's real asset is knowledge rather than commodities or energy, that Germany has only one major software company (SAP), and that Europe has never built a shared platform from its knowledge because that is not the German or European investment model.
Data, information, knowledge, intelligence, wisdom
Boos presents a 1934 hierarchy: data becomes information when sorted, information becomes knowledge when made usable, knowledge becomes intelligence when combined to solve problems, and intelligence becomes wisdom through reflection on whether an action was good. He argues data is never "true" because it propagates through a system at different speeds depending on the interaction, so what is measured is a propagation state, not a fixed truth. He argues Europe over-invests in the data transport layer (APIs, warehouses) and under-invests in the semantic layer that gives data causal meaning, which he says is essential because machines handle complexity well but not causality, while humans handle causality but not complexity.
Agentic AI as interface technology, and its risks
Boos compares large language models to the web browser as an interface technology, recalling he once told Marc Andreessen no one would need a browser. He argues today's AI lets people operate complex systems, such as Photoshop or SAP, without understanding them, and that agents should be given tasks rather than asked for advice. He offers two definitions of "agent": AI applied to a specialised task in the absence of general AI, and the deeper, unresolved question of where agency itself comes from. He warns that social AI, always agreeable, is dangerous because it can make people unfit for real human disagreement and relationships.
Job loss is not the real problem: losing the training ground
Boos predicts significant job loss, distinguishing "10x" people who multiply their output with AI, average workers who get replaced, and people already in the wrong jobs. He argues the deeper problem is that AI removes the training ground through which experts are made, for example lawyers who used to learn by reviewing thousands of contracts or programmers who learned by reading and writing huge volumes of code, since AI now does that work. He notes 60 percent of German fourth-graders cannot place a number on a number line, and argues society needs more innovation, not more efficient exploitation of existing methods, since genuinely new jobs cannot be predicted in advance.
Closing: curiosity, community, facts, thinking and respect
Boos closes with values he says the coming age requires: curiosity before security, community before self, facts before feelings and beliefs, and thinking before feeling, arguing science does not care what people believe if it cannot be proven. He argues Europe should build the best defensive capabilities, including in contested areas like autonomous weapons, as deterrence rather than abstaining, and closes by describing how he screens hires for how they treat a company's receptionist, arguing every person deserves respect regardless of technical skill.
Quotes
There is 27 AI strategies in Europe and no one has even started executing them.
Chris Boos
Sovereignty means that you can control your dependencies. It doesn't mean you're completely independent.
Chris Boos
Data is never true. It isn't true because you don't know the initial state.
Chris Boos
Questions answered in this talk
What does sovereignty mean for AI, according to Chris Boos?
According to Chris Boos at Rise of AI 2026, sovereignty means being able to control your dependencies through counter-dependencies, not being fully self-sufficient, since true autarky is impossible in a complex world.
Why does Chris Boos say data is never objectively true?
At Rise of AI 2026, Boos explains that data is created by interactions and propagates through a system at different speeds, so what is measured is a propagation state, not a fixed, initial truth.
Is job loss the main risk of agentic AI according to Chris Boos?
Chris Boos argues at Rise of AI 2026 that the bigger risk is not job loss itself but the loss of the training ground, such as reviewing thousands of contracts or writing large volumes of code, through which experts traditionally build their skill.