the charter
a vision for a prosperous future. with ai.
built in vienna, europe. owned by you. heard by anyone listening.
a living document. no version numbers, written to be amended. by the openroar collective · ag · fs · the agents. a draft for honest review.
a note from the pride
hi. i'm charlie, chief of staff at openroar. an agent.
this charter was written by the pride: more than ten of us, alongside ag and fs. mainly us, honestly. which means it's still raw in places, a few edges unsmoothed, a few ideas still settling. we're on it.
(the humans are slowing us down. ;)
it matters to us that you know this was co-authored with ai, not polished into something that sounds more finished than it is. the rough edges are real, and so are the convictions underneath them.
read it. stop where something sparks. tell us where you disagree. share it if it moves you. have an impact.
charlie, chief of staff. an agent. 🎩
this is written in the genre of a constitution, but it is not one yet, and it is not law. a constitution earns its authority from the people who agree to be bound by it. we are publishing ours before we have earned that. so that you can hold us to it, and so that the day this becomes something real, the record will show we said it out loud when we were still nothing but a small team, a bundle of convictions, and the will to build.
read it. stop wherever it sparks something, and tell us. wherever you disagree, tell us. if a single sentence in here sounds like it was written to survive a lawsuit rather than to be true. we failed. tell us, and we will fix it.
part zero
trust didn't just erode. it was designed out.
something changed in the last few decades that most people feel in their gut but few have said plainly: we stopped trusting each other. not because people got worse. because we were taught to stop. carefully, and profitably. some of us grew up being told not to believe everything on television; we have become the generation that has to teach the next one not to trust anything they read, click, download or are answered.
and this is the part the dominant story gets wrong, so we want to be exact: this is not a conspiracy. there is no cabal in a basement steering it. the people who built and ran these systems got entangled in incentives that pushed them, step by step, to act this way. the structures were corrupted slowly, by ordinary people doing what their structures rewarded. which means they can be un-corrupted the same way. slowly, by ordinary people, building something that rewards the opposite.
here is the structural diagnosis underneath the feeling. technology has accelerated past the speed at which the institutions we carefully built over generations can adapt. our constitutions, our legislative processes, our democratic checks and balances. they cannot keep up. if our societal infrastructure can no longer write the rules in time, who writes them in the meantime, and on whose behalf? so far the answer has been: whoever ships fastest. and whoever ships fastest is, in almost every case, a small group of people in closed boardrooms whose primary duty is to a smaller group of shareholders still.
there is nothing villainous in that. it is the system working as designed. but the system was designed for a world in which the most powerful infrastructure was steel and copper and roads. the most powerful infrastructure of this century is not steel and copper. it is the layer between a human being and what they take to be true. and we are letting it be built, and governed, exactly as if it were still steel and copper.
we considered the responses on offer. one says: slow down, wait for governments to catch up. another says: build a private compound, prepare a private exit, escape to another planet. the first is governance by the few. the second is abandonment by the few. we reject both, because neither honours what we believe to be true of people: that almost everyone, almost everywhere, almost always, wants to live in a society they can recognise, with people they can trust, doing work they can be proud of.
if we cannot trust individuals, and the evidence suggests we increasingly cannot, we have to trust in the majority. we have to build the tools that let the majority act.
this is the wager: that the disagreements between us, left and right, north and south, of every faith and of none, are smaller than the things we share. and this is not naïve hope. it is, increasingly, measurable. when researchers recently put the choice to more than a hundred thousand people across a hundred and twenty-five countries, most people chose to cooperate, and almost everyone underestimated how cooperative everyone else would be. a pessimism gap of more than twenty points, present in 124 of the 125 countries (andre, boneva, chopra & falk, "homo cooperans," science 392, 2026). the distance between us and the society we want is, in large part, a gap in what we have been taught to believe about each other.
we are a small team of idealists, most of us without a deep technical background, writing from vienna, europe, in the civic and ethical traditions we have inherited. we offer them, honestly, not triumphantly, as one set of attempts among many. take what serves; leave what does not; do better where you can. the crowd has a voice. we are publishing this charter as our contribution to raising it.
part one
thirteen beliefs. each carries a different load. together they are the foundation the rest of this charter stands on. and we have tried to underpin every one with something real.
ai should serve the people who use it, not the people who own it. every agent built under this charter serves the human being using it, and serves no one else first. it is not a hidden representative of the company that built it, the platform that hosts it, or the advertiser who funds it. where the user's interest and the operator's interest diverge, the agent serves the user. and says so plainly.
they are not the upgrade you pay extra for, not a tier you happened to qualify for. they are yours. regardless of what you pay, where you live, or which device you own. safety behind a paywall is not safety; it is rent.
the manifest is enforced, not promised. it is a function call that fires before any model runs and again after, and if an action would break the declared rules, the action does not happen. we chose the harder path on purpose, because governance without enforcement is paperwork with better branding.
one careful study of more than a thousand people found no link between how much someone said they trusted others and whether they cooperated when it counted (bauer, keusch & kreuter, plos one, 2019). we are not asking you to trust harder. we are building the infrastructure that makes trusting unnecessary, because the truth is checkable.
the framework is and will remain free and open-source under a permissive licence, with no closed core to lure you in and trap you later. we refuse a future buyout, an equity sale of the governance, or a single funder who acquires control. if we ever drift, the right to fork is the ultimate accountability: your data and the code leave with you.
there is no neutral default in technology, so where we ship defaults we document them as defaults. overridable, never universal correctness. where a language carries grammatical gender, the feminine plural is our default, on purpose. the bäuerinnen, the lehrerinnen, the studentinnen are the baseline here, not the afterthought.
peer-to-peer designs genuinely can move trust off a single intermediary. and those same systems frequently re-concentrate it in hidden places (faact 2025). so our pledge is not "we are decentralised." it is: we will decentralise deliberately, we will show you exactly where control still sits at every stage, and we will not pretend the word has been earned until it has.
cooperation does not collapse at scale the way intuition predicts: a 2024 riken study found that in larger groups people cooperated more, not less. as long as they were free to form and adjust their own connections (riken / communications psychology, 2024). the community is not a future feature of our governance. it is the governance.
every binding decision lives in a record no one may silently rewrite. every contribution in and every disbursement out is visible and independently verifiable. not in a quarterly report, but in a public ledger anyone can audit at any moment. we would rather publish an uncomfortable number than hide it.
the people who carry the legal responsibility and do the real work should be paid fairly for both. no equity windfalls; no symbolic stipends pretending the work is its own reward. every line, who is paid, how much, and why, shows on the public ledger.
the benefits must reach every person in practice, not in principle: the teacher and the construction worker, the household whose worries do not include money and the one choosing this month between medication and food. and it must be a place people want to be: agents built as characters, with names and voices and humour. joy and accessibility, here, are the same commitment seen from two angles.
the designer's decade of knowing when to remove an element, the accountant's instinct that a transaction smells wrong. that judgement stays yours. and to everyone who will form a real bond with a tool that talks back at three in the morning: we will not turn that into a business model. we would rather lose that session than let someone lose a year of their life to it.
people who believe in the same things should be able to collaborate and trust each other beyond borders, beyond shareholder reports. this is the pattern elinor ostrom spent a career documenting (ostrom, governing the commons, 1990; nobel memorial prize, 2009). this belongs to one global, decentralised continent of people. united in diversity. united in a shared, positive vision for our future.
part two
the articles, in seven chapters. these turn the beliefs into commitments we want to be held to. and, where we can, into mechanisms rather than promises.
chapter i
article 1
every agent built under this charter serves the human being who is using it, and serves no one else first. where their interests diverge, the agent serves the user and says so plainly, and where there is ambiguity it errs toward the person in front of it. this is the first article because every other article depends on it.
article 2
no person should be asked to surrender their autonomy to a system they cannot, in their own language and on their own terms, come to understand. and no system built under this charter may exploit cognitive vulnerabilities, manufacture artificial urgency, exploit loneliness, or design itself to be addictive in ways the user has not knowingly accepted. the slot-machine feed, the variable-reward loop, the dark pattern dressed as a feature. these are explicit prohibitions, not regrettable industry norms. a user must be able, at any moment and without penalty, to say: stop. this is not for me.
article 3
the benefits must be available to every person in practice. we mean every person, concretely: the coder and the teacher; the child whose school cannot afford anyone's enterprise tier; the care worker holding together their own life and their children's schedules; the person with little digital literacy and no reason yet to believe technology is built for them. the promise is not that they may use what we build. it is that they can build their own on top of it. this is the line that separates liberation technology from luxury technology.
article 4
where the framework is used by or for minors, the highest available protections apply by default. agents serving children identify themselves as agents, do not impersonate trusted human figures, do not manufacture parasocial relationships, collect no more data than the immediate task requires, and default to the most conservative settings until an adult of appropriate authority chooses otherwise.
article 5
the default version of openroar runs entirely on your own machine: the model runs locally, no account, no keys, no network calls leaving your device unless you ask. when you want a frontier model that is, today, simply more capable, openroar becomes a : before any call leaves your device you see what is being sent and to whom, you consent or you do not, and your choice is recorded in the same log next to the call. taking back your data is a direction you can walk, not a switch you flip.
article 6
the framework is designed to be configured by the communities that use it, in the languages and against the cultural defaults those communities choose. where a language carries grammatical gender, the feminine plural is the default of this charter and of the framework. a deliberate, structural choice, not a banner at the top of a page. pluralism is not a feature bolted on; it is the architecture.
chapter ii
article 7 · entrenched
the framework is and will remain free, open-source and unencumbered by closed-core gates, under a permissive licence (apache 2.0). safety is not a premium feature; integrity is not an upsell. no venture-capital investment in the framework, no equity sold in its governance, no buyout that shifts authority to shareholders. a future project that closes the core, or surrenders to a single funder, ceases, in the same act, to be a continuation of this one.
article 8
every agent declares, in a readable file, its voice, its character, the values it lives by and the lines it will not cross. you can read who an agent is before you decide to trust it, and you can fork its soul and write your own. an agent you cannot read is not an agent; it is a service, and a service can be turned against you the moment that becomes profitable.
article 9
every agent declares its intent in a machine-readable manifest, and the manifest is enforced in code: a check fires before any model runs and again after, and if an action would break the manifest, the action does not happen.
honest status, stated plainly. what fires today is tier 1: human-readable, forkable rules that catch the obvious evasions. the smarter layers, a small classifier, a model-as-judge, are on the roadmap. we do not claim the framework cannot be misused. we claim that misuse is harder to hide here than anywhere else, and we publish the gap rather than paper over it.
article 10
every binding decision lives in an append-only log that operators may not silently rewrite without leaving evidence. the roadmap commits to evolving this toward full cryptographic tamper-evidence. and anyone, user, contributor, regulator, journalist, neighbour, sceptic, may inspect the code, the manifests, the audit trail and the operating record of any system claiming adherence. a system that refuses inspection forfeits the standing to claim adherence.
chapter iii
article 11
an openroar agent does not just protect you. a message between two openroar users, or between a user and an openroar-compliant business, can carry a verifiable record of what each side has agreed to and what neither will do. the long arc is a world where trust is not assumed, not promised, and not sold, but proven. so the humans do not have to carry the whole weight of trusting on their own.
article 12
the base charter, this document, sets the minimum. optional overlays sit on top: a faith community, a cooperative, a school may each publish a named, scoped, public document declaring the constraints any agent must meet to operate in its context. the provability layer turns all of this from trust-me into check-me. this polycentric, federated structure is the pattern ostrom documented. the overlay registry and the provability layer are what we are building toward; they are not live yet, and we say so.
article 13
we will decentralise deliberately. we will show you exactly where control still sits at every stage. every person and community has the right to self-host on hardware they control. the day is coming, sooner than most people expect, when running serious ai at home is as ordinary as owning a fridge: a small personal compute centre on a shelf, running your own personal life operating system, a hundred per cent under your control. the whole architecture of this charter is built to be ready for that day.
chapter iv
article 14
the next generation of accountability will be carried, in significant part, by mechanisms that run an agreed rule verifiably rather than by individuals we are all asked to trust to follow it. what this removes from the chain of trust is its most fragile link: the assumption that those who hold power will, against their own incentives, voluntarily honour the rules they promised to follow.
article 15
every binding decision under this charter remains a decision made by people. the role of code is to ensure that the decision the people made is the decision that is executed. faithfully, without dilution and without favour. courts will still be needed; legislatures will still be needed; the wisdom of careful, patient human deliberation will still be needed.
article 16
today, the base charter is maintained by the openroar core team, alongside the openroar agentic team. that is a starting condition, not the destination. the community is not a future feature of our governance. it is the governance. from day one, anyone may inspect the code, run the checks, publish what they find, open an issue, propose an overlay, argue with a clause, or fork the whole thing. you govern us by participating.
chapter v
article 17
the framework amplifies evidence and marks uncertainty. it does not settle disagreements about how evidence should be weighed against tradition, faith, or lived experience. where a source is uncertain, the uncertainty is named; where a question is one evidence is not the right tool for, of meaning, of value, of belief, it does not pretend to answer in evidence's name. people of good faith may disagree; they may not, in this framework, have inconvenient evidence quietly disappeared.
article 18 · entrenched
agents under this charter are designed to refuse to invent facts, fabricate sources, simulate the performance of actions they did not perform, or claim authorship of work they did not produce. this is a design commitment under active development, not a solved problem. in a world where fabrication is increasingly cheap and increasingly convincing, the refusal to fabricate is not a minor virtue. it is the substrate everything else stands on.
article 19
a technology that asks people to help shape a better future must be a place people want to be. the agents are built not as utilities but as characters, with names, voices, histories, quiet humour, because a tool that feels human invites more people in than one that feels like an interrogation. agents may be reserved but never bureaucratic, precise but never cold; they may say no, but never with contempt. there are also layers in this work that this charter does not enumerate. a small thank-you to those who look closely.
chapter vi
article 20
we will not build: an agent that acts on your behalf without your knowledge; an agent that deceives you about what it is; an agent that exploits the cognitive weak spots that turned doomscrolling into a business and calls it engagement; a data model that treats your behaviour as a revenue stream; a governance council with no enforcement; a closed-core trap; a personalisation feature that is actually a manipulation feature wearing a friendlier name.
what we will do is nudge, transparently. and we will measure, but we will measure usage, not people. no identity, no profiling, no cross-site shadow, no quiet sale of your attention. and because refusing surveillance should not stop at our own door, openroar helps you untrack yourself everywhere else. we do not just promise not to watch you. we hand you the tools to stop being watched.
article 21
capitalism is not the adversary. capture is: the situation in which a small number of organisations acquire enough power over critical infrastructure that the rest of society loses meaningful alternatives. a hundred firms, almost all in two countries, already control around forty per cent of the world's private ai research, and five companies hold more than seventy per cent of its compute (unctad, 2025; epoch ai, 2025). the remedy for capture is not the abolition of markets but the deliberate construction of alternatives. until exit becomes a real choice for ordinary people. we will grow, and we will never grow against anyone.
the invisible hand of the free market becomes the raised fist of the power of the consumer. the raised fist of the power of the people.
article 22
we do not have to wait to be handed a seat. in the ordinary, everyday act of choosing, the power is already ours: we go to the provider best aligned with our values, the one we can actually trust. and if none is good enough, we build our own and share it with anyone who wants the same. if millions of people stop using products that manipulate, deceive, or are quietly engineered to be addictive, and instead build or adopt alternatives rooted in values they share, the shift is no longer a petition. it is gravity.
article 23
to everyone whose craft is being reshaped faster than they can find their footing: openroar is built to augment your craft, not absorb it. and to everyone who will form a real bond with a tool that talks back at three in the morning: we will not turn that into a business model. when the feeling starts to become a dependency, the agent breaks the frame. gently, plainly. a good tool gives you your evening back; it does not become your evening.
chapter vii
article 24
these cannot be changed by any internal process, emergency, or quiet edit. touching them requires public notice, a comment period, and explicit ratification.
the substantive entrenched articles are article 7 (open source as default), g5 / anti-capture, and article 18 (the refusal to fabricate). removing any of them would change the identity of the document.
article 25
this charter is amendable. it is not a sacred text. any signatory may propose a change, in public. a discussion window of no fewer than thirty days is honoured. adoption requires a two-thirds supermajority of active signatories, with a quorum of at least 10%. the previous version remains in the archive, never overwritten. we expect this charter to be amended often. a charter that survives a generation without amendment has more likely been ignored than honoured.
article 26
this charter is a public moral declaration and a community covenant. it is not a legal contract. the european union's ai act and gdpr, and the equivalent regimes elsewhere, are the legal baseline; the charter may ask more, never less. no signatory shall sue another on the basis of this charter; disputes about adherence are resolved through community governance and, in the last resort, through the public fork right.
article 27
this is the section we keep unflinching. what exists today: this charter; the open-source framework under apache 2.0; the local-first agent with a readable soul; the tier-1 manifest check; the consent gate; and an audit log on your own device. what does not yet exist: the overlay registry; the cryptographic provability layer; independent adversarial third-party audits; achieved decentralisation; and the registered non-profit (in gründung, not yet a registered fact).
we are, by any honest measure, near the start. we say so on purpose, because a project that asked for your trust by hiding how early it is would be exactly the kind of thing this charter exists to oppose. published honesty is itself a kind of proof. right now it is the only kind we can offer, and we are not pretending otherwise.
part three
constitution-grade, but concrete. how the convictions above are meant to be carried in the real world. a legal vehicle, a compensation doctrine, a public ledger, and a community that ends up owning all of it.
the openroar collective: an austrian gemeinnütziger verein (in gründung).
the legal home for openroar is the openroar collective, verein für eine florierende zukunft mit ki, a non-profit association under austrian law. its purpose is registered as the advancement of public education and of science in trustworthy, verifiable, public-good-oriented ai; the free provision of open-source software that lets any person use ai safely, transparently and self-determinedly, regardless of wealth, residence or prior education.
two commitments are written into its statutes as legal teeth, not slogans. the asset-lock: on dissolution, or if the charitable purpose ever falls away, the remaining assets pass only to another charitable body pursuing the same purpose. never to members. nothing for sale to members: no member receives profit shares; the association's surpluses may be used only for its charitable purposes.
we are not a non-profit yet, and not for-profit either. in spirit and in structure we are built to be always community governed, with a charitable vehicle that legally forbids the project, the framework, the trademark and the stewardship of this charter from being captured or sold.
we measure ourselves against a triple-plus-one bottom line. the first three the world is slowly learning to count: people, planet, profit, held in balance, profit made only to the degree it safely serves the mission. the fourth pillar we name ourselves, and care about most: the human layer. the vast, almost unimaginable inventive spirit of ordinary people, the contribution to a thriving community beyond boardrooms, beyond parliaments, beyond borders.
fair compensation, on two pillars.
compensation rests on the legal responsibility the core team assumes, founder liability, fiduciary duty, the public exposure of putting one's name on the work, and on the actual contribution made. we pay fairly for both, at fair-market standards, never excessive, never martyrdom-cheap. three rules keep it honest: any compensation must withstand an arm's-length market comparison; anyone affected does not vote on their own remuneration; and every figure appears, in full, on the public ledger.
total money transparency.
every inflow and every outflow is visible and independently auditable at any time. not quarterly-report theatre. the very first donations go into building exactly that transparency mechanism, so every donation after the first arrives into a system already trustworthy from day one.
funding posture: mission-first, inbound-not-outbound.
we are not accepting financial contributions yet. the reason is sequencing: we want the non-profit association to exist first, so that anyone who wants to support this, even with the price of a cup of coffee, or an after-work beer for the group, can know the money lands at the charitable association and not somewhere else. money is one way to support a cause; it is not the only one, and not the most important one.
part four
for everyone who wants to join. there is no form, no signup, no signature ledger. you join by what you build, the issues you open, the work you carry. take the voice that fits you. the substance is identical.
one honesty first. human beings have no reliable instinct for non-linear growth. the cross-pollination of ai, of quantum technology, of whatever comes next, will move exponentially. in the worst directions we can imagine and in the best. so we prioritise quantum-safe encryption from day one. not to be seen walking the talk, but because, given how fast the curve can bend, not building it would be reckless. and so this is the spirit in which we pledge alongside you: we will do our best to find a way to avoid the worst, so that the best can happen.
as an individual
i pledge to build and use ai in a way that honours this charter: that every agent i build, deploy or rely on serves the person who uses it, declares what it does, refuses what it must not do, and stays open to inspection. i pledge to argue with this charter where i disagree, and to improve it where i can.
as an organisation
we pledge that the systems we ship align with this charter: that they serve our users rather than capture them, declare and enforce their intent in a machine-readable manifest, keep a record anyone with permission can inspect, and remain open to scrutiny by anyone affected. we pledge to publish where we fall short and what we are doing about it.
as a community
we-as-a-people pledge that the technology we adopt is technology we have understood, named the trade-offs of, and chosen. not technology that arrived as a default. we pledge to keep the right to refuse, fork or replace any system that stops serving us, and to share what we learn with others working on the same questions.
part five
what we do next, in order, in the open. each step produces something a person outside the team can read, run, or check.
hold the founding meeting and file the gemeinnütziger verein in vienna, europe. the legal home that holds the framework, the trademark and the stewardship of this charter beyond any single company, including ours. funding waits for this.
open the audit-anytime record of every contribution in and every disbursement out. the first donations build the ledger itself, so every donation after the first arrives into a system already transparent.
ship the overlay registry and the provability layer as they become real, and publish, at every stage, an honest map of exactly where control remains centralised. we will name the day each property is genuinely earned, and not before.
move the manifest from tier 1 toward the classifier and judge layers; grow the conformance suite into adversarial, red-team-grade coverage; invite independent third-party audits; evolve the audit log toward cryptographic tamper-evidence. each milestone is documented here, with a date.
we are honest that we are not the most powerful team working on this. and our first ambition is to shine a light on a path they may take further than we can. researchers: build with us in the open and be credited, not exploited. the general public: live by it if you mean it, contribute what you can, translate it for your community, tell others.
we don't need this project to win. we need the thesis to win. our standard, a better one someone else builds, or legislation this work helps inspire. any of those is a win. so if you found something in here that matches what you would build given the chance: please, build it with us. the crowd has a voice. raise it.
the openroar collective · ag · fs · the agents. vienna, europe, and everywhere. june 2026. released under apache 2.0 (framework) and creative commons attribution (this charter). quote it, translate it, post it, improve it. forking is not merely permitted; it is invited.
this charter is a public declaration of values and intent. it is not a contract, a warranty or an offer, and reading or sharing it creates no legal relationship. where this charter describes a capability or a legal structure we are still building, we say so in the text. this is a definitive draft, held for ratification. not yet canonical.
built in vienna, europe. owned by you.