The ten principles
All ten dimensions must be addressed together. The exact mechanisms remain open to independent research, challenge and possible replacement.
The founder’s conviction is that all ten dimensions are essential. That is different from an obligation to accept any specific cooperative, payment formula or technology design unchanged. None of these principles has been adopted as law, and every number below is a research hypothesis, not a proven result.
When systems act autonomously, protection cannot depend on repairing harm after it has happened.
An AI Guardian framework combining protective AI analysis, technical safeguards, monitoring, auditability and meaningful human intervention.
Whether it can be made operational and enforceable in selected participating European deployments by 31 December 2027. No system can guarantee zero harm.
Society needs to decide openly where human authority must remain, rather than letting technology settle it by default.
A contract that determines which decisions remain exclusively human, where AI must be supervised, which activities can be autonomous and which uses are prohibited.
Where those lines should be drawn, and how a contract of this kind could gain democratic legitimacy.
Ethical commitments protect people only when they can be enforced.
Binding ethical rules with meaningful enforcement, including restrictions or withdrawal of operating rights for serious violations.
Which institutions could enforce these rules, and how they would work alongside existing law.
When an autonomous system causes harm, people need to know who answers for it.
A distinct liability framework covering model developers, application and operating-system providers, deployers and robotic operators, according to their actual obligations and control.
How responsibility should be divided along the chain, and how this fits with existing liability rules.
If beneficial intelligence becomes essential, access to it should not depend on privilege or on dependence on one provider.
Universal access to beneficial intelligence, with freedom of provider choice, interoperability and personal sovereignty.
What universal access should include, who would finance it and how interoperability could be secured.
Freedom from compulsory work must not become freedom from intellectual challenge, and security should not depend on a job that may no longer exist.
Keep education, creativity, entrepreneurship, science, relationships and community contribution central to meaningful lives, and investigate AI-financed income as a form of economic security.
AI-financed income is a research hypothesis. Its level, financing and effect on motivation and employment need independent modelling and pilots.
Foundation models were developed using protected works, in cases where appropriate compensation was not negotiated.
A dedicated fund, financed by model providers, that distributes to verified rights holders using book registries, other intellectual-property records and claims mechanisms.
The legal scope, the valuation and the financing formula all require investigation.
Expertise should remain worth building when AI can draw on it.
Authors, scientists, consultants and specialists turn knowledge into identifiable, licensable AI skills. Users choose which experts, methods and sources their AI uses, and authorised use generates royalties for the owners.
How skills are identified, licensed and accounted for, and whether the royalty model works in practice. A licensed skills and royalty accounting demonstration is proposed.
If autonomous intelligence becomes a major productive resource, the value it creates needs a transparent way to reach those who contribute to it.
A settlement mechanism in which commercially productive uses of AI and robotics, including systems operated inside companies, compensate intellectual-property owners, models, applications, AI operating systems, infrastructure and robotics providers, while accounting separately for contributions to collective prosperity.
Whether such a settlement can be made workable and transparent, and how to avoid double counting. An AI value-settlement transaction demonstration is proposed.
If autonomous intelligence replaces part of the labour that funds wages and social security, citizens need another way to share in the upside.
As research hypotheses: a citizens' cooperative with a 50% ownership objective in qualifying foundation-model suppliers, with returns shared equally, or an AI-specific tax-and-distribution mechanism if that proves unworkable; and a 67/33 financing split for social security between AI-generated value and human labour.
All of these numbers are unproven. Economists and legal specialists must test their sustainability and implementation. A citizen-dividend simulation is proposed.