A Nordic-founded global think-do-tank.
Founded at Harvard and MIT, ImpactLab is a pan-Nordic and global think-do-tank that brings named voices to the same table.
Founded at Harvard and MIT, ImpactLab is a pan-Nordic and global think-do-tank that brings named voices to the same table. We work across the Nordics, the United States, and the Middle East, driving cross-sector collaboration and business-leadership accountability on responsible AI.
A small core. A wide network. Editorial independence enforced by structure, not by statement.
A world where technology serves humanity.
A world where technology serves humanity — building a future that is ethical, sustainable, and inclusive by design.
That sentence is only useful if it is operationalised. For ImpactLab, it means three concrete claims about how AI gets built and deployed over the next decade.
First, that responsible AI is public infrastructure, not a marketing layer. The AI deployed inside hospitals, ministries, schools, and small businesses must be the AI those institutions can defend in front of a citizen, a parliament, and a press corps.
Second, that the test of a responsible AI system is not the press release at launch but the artefacts that remain in place once the launch team has moved on — the procurement clauses, the oversight templates, the literacy modules, the documentation that survives staff turnover.
Third, that the public interest scales when the rooms making decisions include the people the decisions are made for. Inclusion is structural, not stylistic.
Aligned with OECD AI Principles · 2024
Practical, scalable, responsible AI — co-created.
To make responsible and inclusive AI practical and scalable — by connecting leaders, innovators, and communities to co-create actionable blueprints that ensure technology advances the common good.
In practice, three lines of work carry the mission.
Research and writing. The Dispatch is ImpactLab's editorial letter on responsible AI. It is not schedule-bound; we publish when there is an argument worth publishing. It is footnoted, free to read, no advertising, no sponsored content. We publish op-eds and feature interviews in Dagens Næringsliv, VG, TV 2, E24, NRK, and Kapital, averaging more than three placements per month over the last year, alongside contributions to the Harvard Kennedy School blog and MIT Sloan publications.
Programmes. AI for Norway — KI for Norge in Norwegian — adapts Harvard and MIT executive AI content for one million Norwegians, with a six-module curriculum and a 2026 pilot in Østfold. We also run Norwegian-language explainers and policy briefs translating frontier AI research for ministry, municipal, and SME audiences.
Convenings. Closed-door working rooms in Oslo, Cambridge, New York, and the Gulf. Past hosts and partners include Harvard Kennedy School and MIT Sloan. Each room produces a named, implementable artefact rather than a brochure.
We do not implement AI for clients. We do not take vendor commissions. We do not accept sponsored editorial.
What we hold ourselves to.
Blueprints, not whitepapers.
Every engagement ends with an artefact a participant can implement the week they return — a clause, a checklist, a curriculum module — not a 60-page PDF that sits on a shelf. We publish blueprints under open licences where the room permits, and as named-commitment versions where confidentiality is required. Negative results are written up with the same prominence as positive ones.
Named voices at the same table.
Ministers and machine-learning engineers, regulators and founders, civil-society leaders and capital allocators. The decisions that shape AI are made across these boundaries; the rooms must be too. Every ImpactLab convening publishes its participant categories and cohort composition, even when names remain off the record.
Editorial independence.
We accept aligned capital and decline conflicted capital. We publish our funders above NOK 100,000 / USD 10,000 annually. We do not co-brand commercial AI products. The Dispatch carries no advertising, no sponsored content, and no paid placement, and our Advisory Council — forming, with its first meeting in Q3 2026 — has a standing mandate to review editorial independence each year.
Inclusion is structural, not stylistic.
Every Lab and Summit publishes its cohort composition. Equity targets are set in advance, not described after the fact. The same discipline applies inside our published work: vendor selection, training-data documentation, and oversight design are where inclusion is enforced — diversity is downstream of those choices, and we treat it that way.
Nordic in voice, global in reach.
We are Nordic-founded and Nordic-anchored, working in the United States and the Middle East because the responsible-AI questions worth answering are global. We do not export Nordic exceptionalism — we export Nordic operating craft, tested under the EU AI Act and translated honestly into other regulatory and political contexts.
How rooms become commitments become outcomes.
Responsible AI scales when three layers reinforce each other: literacy at the operating level, procurement at the institutional level, and governance at the regulatory level. If any one of the three is weak, the others fail. ImpactLab's theory of change is the operating practice that connects them.
We bring named voices to the same table in a closed-door working room around a single concrete question. Rooms are small (10 to 30 people), bilingual where useful, off-the-record by default, and structured around a question that has a defensible artefact at the end of it. Past convenings have been hosted in Oslo, Cambridge, New York, and the Gulf; past hosts include Harvard Kennedy School and Nordic public-sector partners.
In 90 minutes, the room produces an artefact — a procurement clause, a model-card template, an oversight protocol, a curriculum module — that participants commit to implement inside their own institutions. Co-authorship is the discipline: an artefact is owned by everyone in the room or it is not the right artefact.
Participants take the artefact home. ImpactLab publishes the open-licensed version, or the named-commitments version where confidentiality is required, and follows up at 90 days, 6 months, and 12 months to record what was implemented, what was modified, and what failed. The follow-up record becomes the next Dispatch issue, the next blueprint, and the input to the next room.
The geographic logic mirrors the operating logic. The Nordics are where small-state regulatory craft can set a defensible national pattern before EU AI Act enforcement begins on 2 to 3 August 2026. The United States is where research and capital concentrate and where assurance standards are set. The Middle East is where state-led AI deployment is happening at the largest scale. A pattern that works in one region, written up honestly, often works in another. That is the leverage.
Independent. Norwegian-registered. Transparent by default.
ImpactLab is a Norwegian non-profit association (ideell forening) registered with Brønnøysundregistrene. Organisation number 835157032. We operate as a public-interest think-do-tank. Headquartered in Oslo, with operations in Cambridge, Massachusetts (C10 Labs at 1 Broadway) and a working presence in New York. The organisation operates with a small permanent core and a working network of more than 200 contributors across the Nordics, the United States, and the Middle East.
Forming — first meeting Q3 2026. Founding members announced with the meeting agenda. The Council's standing mandate is to safeguard editorial independence on The Dispatch, advise on programme selection, and review partnerships above a defined threshold. A code of conduct, conflict-of-interest policy, and gift-and-honoraria policy will be published when the Council is named. Full governance detail is on the governance page.
Our work is editorially independent of our funders, partners, and sponsors. Conflicts of interest are disclosed in every published artefact. The Dispatch does not run advertising, sponsored content, or paid placement; ImpactLab does not co-brand commercial AI products; and we publish all funders above NOK 100,000 / USD 10,000 annually.
Each year we publish a transparency report listing all funders above the disclosure threshold, all partner institutions, all blueprints produced, and a self-assessment against our Principles. The 2026 report publishes in February 2027 and will be archived at /governance.
Building blueprints for responsible AI — Nordics · US · Middle East.