Think tank · Policy · Basel, Switzerland (with an EU office in Brussels and an office in Bern as listed by Pour Demain).
Pour Demain is a Swiss- and Belgium-registered non-profit think tank that works at the intersection of science, policy, and civil society. Its stated mission is to develop evidence-based, targeted policy proposals for pressing societal risks—especially advanced AI safety and governance, and biosecurity/pandemic preparedness—with an explicit focus on translating recommendations into actionable strategies for policymakers and broader public understanding.
In AI safety, Pour Demain focuses on practical governance design and implementation for Europe’s evolving AI regulatory framework (including the EU AI Act’s operationalization and related compliance mechanisms). Its work emphasizes mechanisms such as third-party/independent evaluations of advanced models and safety-related incident response. In biosecurity, Pour Demain (notably “from 2022 to 2025” per its own site) advances policy thinking around biosafety/biosecurity readiness, including pandemic preparedness and governance approaches for tractable near-term risk-reduction opportunities such as nucleic-acid synthesis screening. Strategically, Pour Demain positions itself as an enabling civil-society actor for European institutions and cross-stakeholder processes—providing technical/policy input, convening expert audiences, and publishing briefings and recommendations directed at EU and Swiss decision-makers. It also reports geographic expansion via a Brussels office to extend its remit beyond Switzerland.
No curated programs yet. Check the careers page ↗.
Bringing independent oversight inside frontier AI companies Cover image: Stock photograph from Pexels, used under a free licence. Most independent evaluations of frontier AI test models from the outside, through an API, before release. But many of the most serious risks now come
Embedded Assessments for Frontier AIBringing independent oversight inside frontier AI companies Cover image: Stock photograph from Pexels, used under a free licence. Most independent evaluations of frontier AI test models from the outside, through an API, before release. But many of the most serious risks now come
Public Transparency for Frontier AIEurope’s opportunity to raise the global floor before it’s too late
Public Transparency for Frontier AIEurope’s opportunity to raise the global floor before it’s too late
Pour Demain: Work in Review (2025–2026)Over the past 18 months, Pour Demain has contributed to the implementation of Europe's emerging AI governance framework through policy engagement, technical analysis and stakeholder convening. Alongside AI, we’ve continued to advance evidence-based biosecurity policy and support
Pour Demain: Jahresrückblick 2025–2026In den vergangenen 18 Monaten hat Pour Demain einen Beitrag zur Ausgestaltung und Umsetzung des entstehenden europäischen Rahmens für die Governance von Künstlicher Intelligenz (KI) geleistet – durch politische Mitwirkung, technische Analysen und den Austausch mit Politik, Wissen
Pour Demain: Work in Review (2025–2026)Over the past 18 months, Pour Demain has contributed to the implementation of Europe's emerging AI governance framework through policy engagement, technical analysis and stakeholder convening. Alongside AI, we’ve continued to advance evidence-based biosecurity policy and support
Confidential computing can enable better frontier AI auditingThe first public measurement of confidential-computing overhead on an instrumented evaluation of a frontier AI model shows the cost is manageable and the remaining gap is a solvable engineering problem. Cover image: Stock photograph from Unsplash, used under a free licence. Autho
Confidential computing can enable better frontier AI auditingThe first public measurement of confidential-computing overhead on an instrumented evaluation of a frontier AI model shows the cost is manageable and the remaining gap is a solvable engineering problem. Cover image: Stock photograph from Unsplash, used under a free licence. Autho
Frontiers Article on NA synthesis screeningInnovation needs trust: Why Europe should see biosecurity screening as an enabler of biotechnology Europe's biotechnology sector is entering a period of rapid transformation. Synthetic nucleic acid technologies are accelerating breakthroughs in healthcare, diagnostics and sustain
Frontiers Article on NA synthesis screeningInnovation needs trust: Why Europe should see biosecurity screening as an enabler of biotechnology Europe's biotechnology sector is entering a period of rapid transformation. Synthetic nucleic acid technologies are accelerating breakthroughs in healthcare, diagnostics and sustain
Frontiers Article on NA synthesis screeningInnovation needs trust: Why Europe should see biosecurity screening as an enabler of biotechnology Europe's biotechnology sector is entering a period of rapid transformation. Synthetic nucleic acid technologies are accelerating breakthroughs in healthcare, diagnostics and sustain
Pour Demain summarizes its 2025–2026 activities, including support for Europe’s AI governance framework (e.g., EU AI Act-related work such as serious-incident response, external evaluation mechanisms, and AI Office implementation capacity) and continued evidence-based biosecurity work through early 2026 (including contributions related to nucleic-acid synthesis screening).
Confidential computing can enable better frontier AI auditingPour Demain argues—through a technical policy/technical-auditing lens—that confidential computing/TEE-style approaches could enable better governed access for frontier AI auditing where black-box evaluation has limitations, proposing a mechanism to protect trade secrets while releasing bounded evidence bundles.
Loss of control: closing the governance gap before the first incidentPour Demain discusses “loss of control” governance for advanced AI systems and links it to near-term incident risk framing and oversight/organizational capacity needs (as described in the post).
From Ambition to Action: Equipping the EU AI Office for AI Act EnforcementPour Demain calls for resourcing and capacity-building for the EU AI Office as enforcement scales, outlining roles/units and policy actions to address implementation needs.
Responding to Serious AI Incidents: Policymaker BriefingPour Demain publishes a policymaker briefing focused on AI incidents in the context of the EU AI Act’s Code of Practice for general-purpose AI models and suggests actions for policymakers over immediate, medium, and long-term horizons.