
About DARI
Media Relations
Press contact, boilerplate, positions and media assets for newsrooms reporting on DARI and on humanity's sovereignty in a world shared with superintelligent AI.
DARI is an international non-profit alliance for humanity's sovereignty in a world shared with superintelligent AI.
Based in Geneva · Guiding principle: Being Human Still Matters
Statements and press releases
Current statements, reactions to news events and press releases are published in News & Media, where they can be filtered by topic and downloaded as PDFs.
Press contact
Aleksandar Stojanovic
Founder and Director General
Mobile: +49 …
Email: presse@dari.earth
Interviews in German and English - live, by phone, by video or in writing. We answer newsroom enquiries within two hours on working days, and faster when the news demands it.
Topics you can approach us about
Self-improving AI and superintelligence
Where development actually stands, what is established and what is not.
Who decides?
The democratic legitimacy of decisions of civilisational consequence that are currently taken inside a handful of companies.
Red lines and verification
What arms control and international humanitarian law teach us about governing AI - and where the analogy breaks down.
Human agency
How control is lost not only in one dramatic rupture but gradually: through delegation, dependency and convenience.
Europe and the frontier
The AI Act, oversight of the most capable models, and the options open to Europe.
Everyday AI vs. frontier risk
Why alarmism and dismissal are equally wrong - and why the two must be kept apart.
What DARI stands for - in five sentences
- 01The question of whether superintelligence can be safely controlled is, according to the people whose job it is to answer it, unresolved.
- 02Control is the laboratories’ question. Sovereignty is society’s question: who gets to decide whether this risk is taken at all?
- 03We are not calling for a general AI moratorium, but for a red line at one precise point: no fully self-improving systems without independently verified containment.
- 04Humanity has contained existential technologies before - never voluntarily, never without verification, and almost always only after a shock. This time it has to work beforehand.
- 05Being Human Still Matters: human agency is a protected good in its own right, not a by-product of technical safety.
Downloads
- Press kit (PDF) link
- Position paper “Who decides about superintelligence?” (PDF) link
- Press photos of Aleksandar Stojanovic, 300 dpi, portrait and landscape link - free of charge for editorial use with the credit: photographer name / DARI
- Logo package (SVG, PNG; light and dark) link
Transparency
DARI is independent. insert funding statement from Transparency, variant A or B Our Director General works full time in AI consulting. We disclose this because newsrooms should know it before they find out themselves. Details in the Transparency tab.
Boilerplate
Approved standing description
One sentence - for captions, lower thirds, bylines
DARI is an international non-profit alliance, based in Geneva, for humanity's sovereignty in a world shared with superintelligent AI.
Three sentences - for news items and article endings
DARI is an international non-profit alliance, based in Geneva, for humanity's sovereignty in a world shared with superintelligent AI. It develops norms, foresight programmes and public mandates so that people remain a party to their own future, even alongside systems that exceed them cognitively. Its initiatives include the Human Agency Deterrence Doctrine, the DARI Pact, the SuperFutures foresight programme and the Tesseract archive of human creation.
One paragraph - for background pieces and event programmes
DARI is an international non-profit alliance based in Geneva. It starts from a sober observation: several companies are avowedly working towards AI systems that surpass humans in nearly all cognitive capabilities - and the decision to take that risk is so far being made without a democratic mandate. DARI works on the conditions under which humanity remains the subject of its own future even in such a world: with a normative framework for human agency (Human Agency Deterrence Doctrine), a public mandate (DARI Pact), a foresight programme for decision-makers and the public (SuperFutures) and an archive of human creation (Tesseract). Its advisory board includes Yves Daccord, former Director-General of the International Committee of the Red Cross, and Vaughn Tan (University College London). DARI was founded and is led by Aleksandar Stojanovic. Guiding principle: Being Human Still Matters.
Factsheet
Mission and approach
The starting point. The leading AI companies are openly working towards systems that surpass humans in nearly all cognitive capabilities. Whether and when they will succeed, nobody knows. Two things are established: the safety researchers at these companies say publicly that they do not yet have a robust plan for reliably controlling such systems. And the decision to take the risk is being made inside a few companies, under competitive and capital-market pressure, without a democratic mandate.
Our question.The technical question - how do you make AI safe? - belongs to the researchers. DARI asks the question that comes before it: under what conditions does humanity remain the subject of its own future? Who gets to decide which risks are taken on everyone's behalf? And what has to be protected before it is too late to negotiate over it?
Sovereignty, not only control. Control describes the relationship between a laboratory and its system. Sovereignty describes the relationship between a society and its future: the ability to set its own conditions and to refuse those of others. Technical containment is necessary, but it does not replace the question of who decides on deployment.
Why Geneva. Geneva is the place where humanity learned to set limits on itself - with the Geneva Conventions, the Conference on Disarmament and a tradition of humanitarian diplomacy. That tradition shows what works: norms before the technology outruns them; verification instead of trust; the right to say no. It also shows what fails: prohibitions without oversight remain declarations of intent.
Coexistence is not capitulation. DARI works from the assumption that people may in future have to share the world with intelligence greater than their own. That does not mean declaring it inevitable or simply accepting it. Coexistence means the conditions are negotiated - and that includes the ability to limit or refuse particular developments.
The initiatives
Human Agency Deterrence Doctrine
A normative framework that anchors human agency as a protected good in its own right in policy, law and institutional design. The doctrine transfers the logic of deterrence to humanity's relationship with superintelligent AI: it sets out conditions and safeguards intended to make overriding human decision-making authority untenable for states, companies and systems alike. ⚠ reconcile with actual doctrine content Status: in development; version 1.0 planned for date.
DARI Pact
A public commitment by people and organisations, insert core sentence of the Pact. So far 600+ individuals and organisations have signed. The Pact is the public mandate DARI invokes in its dealings with policymakers and institutions.
SuperFutures / FutureScapes
A programme for futures literacy, foresight and deliberation: decision-makers, institutions and the public think through possible AI futures in a structured way before those futures arrive - with scenarios, formats for joint deliberation and publicly accessible programmes. The programme entity is a non-profit GmbH under German law; the academic partner is the University of Verona.
Tesseract
An archive of human creation on one million memocubes made of quartz glass, designed for very long timescales. ⚠ specify time horizon and archive contents Tesseract records what people have brought forth - as cultural memory and as a reminder of what is at stake.
How DARI intends to have an effect
DARI pursues three pathways of impact, each supporting the others:
Norms. The Human Agency Deterrence Doctrine is intended to establish human agency as a distinct object of governance - citable in political processes, transferable into regulation, procurement standards and institutional design.
Foresight and deliberation. Through SuperFutures, decision-makers learn to play through possible AI futures before those futures create facts on the ground. The aim is demonstrably changed decisions, not slide decks.
Cultural permanence and public mandate. Tesseract and the DARI Pact give the cause a public foundation: an archive of what is to be protected, and a growing number of people who stand behind it.
Core positions
- 01
The problem is unresolved - according to the people meant to solve it.
Leading safety researchers at the frontier laboratories say publicly that there is no robust plan for reliably controlling superintelligence. This is no longer a fringe opinion; it is the state of the industry.
- 02
Control is the laboratories’ question. Sovereignty is society’s question.
Whether a system can be technically contained is an engineering question. Whether the risk may be taken at all is a political one. At present it is being answered by companies in competition with each other.
- 03
A red line at one precise point - not a blanket moratorium.
DARI is not calling for a ban on AI. We are calling for systems that develop their own successors largely without human control to be neither trained nor deployed until it has been independently demonstrated that they can be contained, monitored and shut down. The burden of proof lies with the developers.
- 04
Everyday AI is not the problem.
The AI assistant at a local government office and the frontier laboratory working on self-improving systems are two fundamentally different risk classes. Conflating them paralyses useful applications and trivialises the actual risk.
- 05
Bans without verification are declarations of intent.
Humanity has contained existential technologies - but only effectively where verification existed. For AI that means compute capacity, training runs and evaluations must become independently auditable.
- 06
Agency is usually lost gradually.
It is not only catastrophic loss of control that threatens human self-determination, but the quiet delegation of decisions to systems without which the state, the economy and daily life eventually cease to function. Critical institutions must be able to demonstrate that they remain operational without AI.
Approved quotes
All quotes: Aleksandar Stojanovic, Director General, DARI
⚠ to be personally approved before use
“The story this week is not that somebody resigned. The story is that the people whose job it is to make superintelligence safe are saying publicly that they do not yet have a plan for it.”
“Control is the laboratories’ question. Sovereignty is society’s question.”
“We do not need a moratorium on AI. We need a red line at the point where machines begin building their own successors.”
“The AI assistant at the local government office is not the problem. Lumping it together with superintelligence trivialises the actual risk.”
“Geneva is the place where humanity learned to set limits on itself. We need to learn that again - and this time before the shock.”
“Agency is rarely lost with a bang. Usually it is lost because delegating is more convenient.”
“If leading safety researchers consider a catastrophe to be double-digit likely, any other high-risk industry would stop. In AI, we accelerate.”
Pointed - only use if you can defend it live.
Spokesperson: Aleksandar Stojanovic
Founder and Director General
Short bio - one sentence
Aleksandar Stojanovic is the founder and Director General of DARI and has worked for some 25 years as an entrepreneur and manager in the European technology industry.
Short bio - three sentences
Aleksandar Stojanovic is the founder and Director General of DARI, an international non-profit alliance based in Geneva that works for humanity's sovereignty in a world with superintelligent AI. Across some 25 years in the European technology industry he founded the digital agency Digital District (sold to Publicis), led Razorfish Germany as managing director, and co-founded AVA ⚠ description needed, which raised 12 million euros and was sold in 2022. His main occupation is Head of AI at Sopra Steria Next Germany; at DARI he combines the practice of AI adoption with the question of where its limits must lie.
Long bio
Aleksandar Stojanovic is the founder and Director General of DARI. He has worked for some 25 years at the intersection of technology, company building and transformation. He founded the digital agency Digital District and sold it to Publicis, was managing director of Razorfish Germany and co-founded AVA ⚠ description needed, which raised 12 million euros in capital and was sold in 2022. In 2024 and 2025 he was VP Strategy at the Berlin AI company Merantix and CEO of Tendra, an agentic-AI venture for tendering and procurement processes.
Today he heads the AI business of Sopra Steria Next Germany as Head of AI across all sectors, with a focus on regulated markets such as financial services and the public sector. He advises the Edgelands Institute ⚠ verify wording regarding the Berkman Klein Center at Harvard University and sits on the AI advisory board of Cornelsen.
At DARI he brings together two perspectives that rarely meet in one person: practical experience of how AI is actually introduced in banks, insurers and public authorities, and the question of where development needs limits that companies can no longer draw for themselves. He lives in Hamburg and Berlin.
Interview topics
Superintelligence and self-improving AI · democratic legitimacy of AI decisions · red lines and verification · human agency · Europe and the AI Act · the difference between everyday AI and frontier risk
Interview languages
German, English further languages
Disclosure
Aleksandar Stojanovic works full time as Head of AI at Sopra Steria Next Germany and in that capacity advises companies and public authorities on AI adoption. As Director General of DARI he speaks exclusively for DARI. DARI's positions are not the positions of Sopra Steria. ⚠ to be agreed with Sopra Steria
Lower third / on-screen credit
Aleksandar Stojanovic · Director General, DARI, Geneva
Advisory board
Yves Daccord was Director-General of the International Committee of the Red Cross (ICRC) from 2010 to 2020. ⚠ add current roles; have bio approved
Vaughn Tan teaches at the UCL School of Management in London and researches how organisations decide under genuine uncertainty. He is the author of The Uncertainty Mindset (Columbia University Press, 2020). ⚠ verify current role; have bio approved
further advisory board members
Stages and collaborations
Stages. Representatives of DARI have appeared at ⚠ substantiate format, year and speaker in each case: DLD, World Economic Forum, Nexus, TEDx India, Web Summit
Collaborations. The academic partner of SuperFutures is the University of Verona. DARI is in exchange with ⚠ specify the nature of the collaboration and obtain approval: Long Now Foundation, Qatar Foundation, Consciousness Foundation, Faculty of Philosophy [University] Munich, Discover Collection
Transparency, funding, conflicts of interest
Funding. Since the beginning of 2025 DARI has been funded essentially by its founder, supplemented by revenue from the Tesseract programme and further sources.
Independence - choose one variant ⚠ decision pending
Variant A - only if fully accurate
DARI has not accepted funds from companies that develop frontier AI models, and will not do so in future.
Variant B - narrower wording
DARI does not fund its position and policy work from companies that develop frontier AI models. Contributions from technology companies to individual programmes are disclosed.
Separation of non-profit and revenue. Commercial activities (Tesseract, FutureScapes commissions) are handled through separate entities ⚠ verify status.
Self-commitment. From 2027 DARI will publish annually all contributions above CHF 10,000 with the names of the donors.
Conflicts of interest. The Director General works full time in AI consulting (see People). Advisory board members disclose their relevant activities on the website.
Questions and answers for newsrooms
What is DARI in one sentence?
An international non-profit alliance based in Geneva that works to ensure humanity decides its own future even in a world with superintelligent AI.
You are not AI researchers. Why should anyone listen to you?
Because the technical question is currently being answered by the researchers themselves - and their answer is: we do not yet know how. Our question is a different one: who gets to decide whether this risk is taken, and under what conditions? That is a question for institutions, international law and the public, not for laboratories. It calls for experience with institutions, negotiation and governance - which is why a former Director-General of the ICRC sits on our advisory board.
Your Director General sells AI adoption for a living. How does that fit together?
That is precisely why he can name the difference. The risks of an AI assistant in a public administration are real but manageable - there are rules, audits and accountable people for that. A system that builds its own successors is a different risk class, for which none of that exists yet. Conflating the two paralyses the economy and trivialises the real problem. Our position: deploy everyday AI responsibly, and draw a red line at the frontier.
Is superintelligence coming? And when?
Nobody knows, and we consider bets on specific years unserious in both directions. What matters is this: the leading laboratories are avowedly working towards it, and their own safety researchers say development is moving faster than expected. Policy that waits for certainty acts too late.
Are you calling for a pause or a ban?
No general pause and no ban on AI. We are calling for a precise red line: systems that develop their own successors largely without human control must not be trained or deployed until it has been independently demonstrated that they can be contained, monitored and shut down. The burden of proof lies with the developers, not with the public.
How high do you estimate the risk of catastrophe?
We do not name a figure of our own, because any figure would feign a precision that does not exist. But we take seriously what the industry itself says. Nuclear power plants are designed for probabilities of severe core damage on the order of once in ten thousand to a hundred thousand reactor years. When leading safety researchers speak of double-digit percentages for AI, any other high-risk industry would stop and prove its case before continuing.
Is warning about superintelligence not hype that serves the laboratories?
The objection is fair wherever warnings serve as marketing. But it does not explain why employees resign and risk their careers in order to warn. And it is not an argument against rules - it is an argument for them: anyone who believes the laboratories are exaggerating should have no objection to binding, independent proof. That proof would expose the exaggeration, or confirm the risk.
Does the superintelligence debate not distract from real harms happening today?
Only if you play them off against each other. Job losses, discrimination by systems and dependency on a handful of providers are real problems - and they are part of the same question: how much decision-making power do societies give up, and to whom? Human agency is the bracket that holds both together.
What is DARI’s position on Anthropic, OpenAI and the other laboratories?
We do not attack individuals or single companies. The problem is structural: even laboratories that take safety seriously are in a race in which stopping costs market share. Self-commitment is therefore not enough. What is needed are rules that apply to everyone at the same time.
What should Germany do concretely?
Three things: push within the EU for the obligations on the most capable AI models under the AI Act to cover loss-of-control scenarios and published emergency plans; build sovereign state capacity to audit frontier models independently rather than relying on self-reporting; and initiate international talks on the verification of frontier compute capacity.
Who funds DARI?
See Transparency. In short: so far essentially the founder, plus programme revenue. insert independence statement, variant A or B
What is the Human Agency Deterrence Doctrine?
A normative framework that establishes human agency as a protected good in its own right and sets out conditions under which overriding human decision-making authority becomes untenable for any actor. The doctrine is in development; version 1.0 is due ⚠ reconcile with actual doctrine content; publication date
Photography and logo
Press photos
- Portrait of Aleksandar Stojanovic, portrait format, neutral background
- Portrait of Aleksandar Stojanovic, landscape format (for online leads, 16:9)
- Situational photo, stage / talk
- optional: memocube object photo (Tesseract)
- optional: advisory board / team
All photos 300 dpi, at least 3000 px on the long edge, free of charge for editorial use with the credit: photographer name / DARI.
Video
- Interview b-roll, broadcast quality link
- Foundation showreel, 16:9 and 9:16 link
- Keynote and conference footage link
Free of charge for editorial use with the credit: DARI Foundation. Raw footage on request.
Logos and press kit
- Logo package, SVG and PNG, light and dark link
- Press kit, complete (PDF) link
- Factsheet and boilerplate (PDF) link
Caption template
Aleksandar Stojanovic, founder and Director General of the Geneva-based AI alliance DARI. Photo: photographer name / DARI
Logo
SVG and PNG, for light and dark backgrounds, with clear space. Do not distort, do not recolour.
Contact
DARI
postal address, Geneva
dari.earth
Press contact
Aleksandar Stojanovic, Founder and Director General
Mobile: +49 …
Email: presse@dari.earth
Interviews in German and English · reply within two hours on working days
Requests
Interviews, statements, keynotes and appearances
We answer newsroom enquiries on working days within two hours, and faster when the news demands it. Interviews are available in German and English - live, by phone, by video or in writing.
Interview
On or off camera, in studio or remote. Our Director General is the primary spokesperson; advisory board members are available on specific subjects.
Statement or expert comment
A written position on a news event, usually within the same working day. Attributable quotes supplied ready to publish.
Keynote or panel
Conferences, policy forums, corporate and academic events. Topics span superintelligence, governance, red lines and human agency.
Background briefing
Off-the-record orientation for journalists preparing a longer piece, with access to our research and scenario material.
Other appearances
Podcasts, documentaries, workshops and editorial collaborations.
Send a request
⚠ form is a placeholder - not yet wired to a handler