Who Checks the AI? Rebel AIs, Agents and Doing Better Otherwise | Slush Side Event
A Slush 2026 side event: an afternoon with AICRIT (University of Turku) on critical AI, independent oversight and the agents that will decide next. Free of charge, lunch included.
About the event
The world is being rebuilt on opaque AI systems that take part in decisions about credit, insurance claims, benefits, permits and public services. We are told these systems are neutral. They are not. Most are “black boxes” that concentrate power, blur accountability and amplify systemic bias. Relying on Big Tech to police themselves is a vulnerability. This is a structural threat to democratic governance, scientific integrity, and planetary well-being.
AICRIT (University of Turku) is a multidisciplinary research project on how artificial intelligence and power, knowledge and governance shape each other. It aims to democratize AI and support a more sustainable global economy, and builds critically proven AI for decision support, analytics, transparency, ethics and sustainability. AICRIT sprang from an Academy of Finland funded research project and has been financed by a research-2-business grant from Business Finland.
Its Second Opinion Protocol puts adversarial agents inside the decision chain: they look for hidden assumptions and bias, trace every claim back to its source, and actively search for the evidence that contradicts a conclusion, not only the evidence that supports it. The result is a decision record people can inspect, replay and challenge. Our starting point is simple: critical thinking is not a prompt; it is an architecture.
The fundamental question is then shifting from whether to use AI to how to show that its decisions can be trusted. In high-stakes work, an AI that only tells you what you want to hear is a liability, and an answer that cannot be traced or cannot be defended to a board, a regulator or a customer. AICRIT's answer is Rebel AI: critical thinking, verification and debate built into the stack itself. In practice that means checking strategies and policy proposals for greenwashing or hidden lobbying, stress-testing AI-generated reports before they are relied on, and leaving a decision trail a human can inspect. The EU AI Act's obligations for high-risk systems apply from December 2027. Where the data stays and how much energy the reasoning burns are becoming part of the same question. This afternoon brings researchers, academics, companies and the public sector together to look at what workable oversight looks like in practice, and how to choose better options on every level of the AI stack. It takes place on the Monday of Slush week.
AICRIT Seminar Program
12.00 Welcome by Organisers: AICRIT as R2B going forward
12.15 Keynote Address Maija-Riitta Ollila
13.00 Lunch (open) / Lunch Discussion (separate meeting room)
13.45 AICRIT / AI Otherwise: Choosing better options on all levels of stack / Outi Korhonen and Jari Ala-Ruona
14.30 Break
15.00-16.00 Panel / Fishbowl Discussion for Future of Rebel AIs, Agents and Doing Better Otherwise
Who should come
Leaders and specialists who deploy or govern AI in decisions that matter, in finance and insurance, the public sector, critical infrastructure and technology companies. Also, researchers in law, ethics, computer science and the social sciences working on AI and accountability.
Registration by 9 November 2026.
More about AICRIT: www.aicrit.fi