Unsovereign AI | What About My Data?

Responsible AI protects our data. Unsovereign AI extends that ambition by design: the machine is prevented from becoming sovereign over the person's inner world, preserving both privacy and human agency.

Responsible AI

“What about my data?” is often the first question people ask when encountering a new artificial intelligence system.

The RIZOM App Privacy Notice provides a concrete answer. 

• Your words and voice belong to you. 

• Personal reflections are stored and processed locally on your device. 

• Apart from the email address required to maintain an account, personal content is not transmitted to RIZOM’s servers unless the user actively chooses to share it. 

• Shared material is anonymised before being aggregated or used for research and insight generation. 

• RIZOM does not use personal data for profiling, commercial targeting or third-party marketing.

These provisions matter. They establish ownership, consent and control in accordance with the UK GDPR and EU GDPR. But compliance is only the beginning. 

The more consequential question is:

Who acquires the power to interpret me?

Privacy is not enough

Much of the current debate treats personal data as an asset that must be stored securely, processed lawfully and shared only with permission.

This remains necessary. Yet data used by artificial intelligence is rarely passive. Words, hesitations, searches, preferences, images, moods and decisions become material from which systems infer who a person is, what they may want and what they are likely to do next.

An AI system does not need to own someone’s data to acquire influence over them. That influence begins when the system becomes the primary authority that names their experience, selects what is relevant, anticipates their behaviour or formulates their response.

The individual may remain legally recognised as the data subject while becoming, functionally, the object of machine interpretation.

This is where the question of sovereignty enters. Sovereignty is not simply the ability to refuse data collection. It is the ability to retain authorship over one’s experience and agency over what follows from it.

The gradual transfer of sovereignty

The prevailing model of generative AI asks the person to externalise a problem, memory, question or unfinished thought. The system processes that material and returns a completed form: an answer, interpretation, summary, recommendation or decision.

This can be useful. It can also produce a subtle transfer of capacity.

When a machine repeatedly performs the work of naming, composing, judging or deciding, the individual may become increasingly dependent on the output. The system does not merely assist an existing human capacity. It can begin to stand in for it.

A service may therefore be privacy-compliant, secure and transparent while still weakening the agency of the person using it.

The issue is not that the machine has become conscious or acquired sovereignty in a legal sense. It is that the architecture has granted it a progressively sovereign role in the relationship. The machine becomes the interpreter; the person becomes the source material.

RIZOM starts from a different premise.

What must remain with the person

RIZOM’s Symbolic Operating System is designed around capacities that should not be surrendered to the machine:

• authorship of one’s experience;

• the freedom to express something without being reduced to a category;

• the ability to interpret and reinterpret what it means;

• control over what is retained, shared or withheld;

• the right to change direction;

• and the capacity to act from one’s own judgement.

These principles are not added as an ethical layer after the technology has been built. They shape the architecture itself.

Personal content remains local to the device. Consent is required before sharing. Data collection is minimised. Personal reflections are not assembled into a central corpus for commercial profiling. Where information is voluntarily contributed to collective insight, it is anonymised and subject to aggregation thresholds intended to prevent re-identification.

This is lawfulness by design.

Policies may prohibit misuse. Architecture can prevent certain forms of misuse from becoming technically available in the first place.

The distinction is important. A promise not to exploit centrally held personal material still leaves the material available for exploitation. An architecture that does not collect that material removes a category of risk rather than merely regulating it.

Yet RIZOM’s approach goes further than data localisation. It also addresses what the interaction itself does to the person.

Recursive, not generative

Generative AI is generally designed to produce something for the user.

RIZOM is designed to return something to the user.

That difference defines the Basho Loop™, the core symbolic engine of RIZOM’s architecture.

Explore the Basho Loop™

The person is invited to express experience through metaphor rather than select from a predefined diagnostic label. Each subsequent prompt is informed by what came before, preserving continuity across time. Metaphors begin to form motifs; motifs form trajectories; and those trajectories make patterns of meaning visible without replacing the person’s authority to interpret them.

The system does not declare: This is what you feel. It helps create the conditions in which the person can say, with increasing depth and precision: This is what my experience means. That is recursion.

The individual contributes an expression. The system reflects its emerging structure. That reflection returns to the individual, who remains able to accept it, reject it, alter it or take it somewhere unforeseen.

Meaning is not extracted and converted into a machine-owned conclusion. It circulates back to its human source. The machine participates in the process without becoming sovereign over it.

A scaffold that is meant to come down

The Basho Loop™ distinguishes between a prosthesis and a scaffold.

• A prosthesis performs a function on someone’s behalf. The more effectively it performs that function, the more indispensable it may become.

• A scaffold supports the development of a capacity. Its success is measured by the extent to which that capacity can eventually stand without it.

  

RIZOM is conceived as a scaffold.

It holds open a reflective space while the person strengthens their ability to recognise patterns, articulate experience and exercise judgement. Its purpose is not to maximise dependency, engagement or the frequency of interaction. Its purpose is to help build a capacity that belongs to the person. 

The scaffold is meant to come down. This creates a very different relationship between human and machine. The machine does not become more sovereign as the interaction progresses. The individual becomes more capable of sovereignty.

Beyond responsible AI

Responsible AI frameworks commonly address essential questions such as safety, fairness, transparency, accountability, privacy and regulatory compliance.

But they do not fully answer another set of questions:

• What human capacity does this system cultivate?

• What capacity does it displace?

• Does using it leave the person more discerning or more dependent?

• Can the person contest the system’s interpretation?

• Does the architecture preserve space for uncertainty, opacity and revision?

• Does the system strengthen the user’s capacity to assume responsibility, or assume responsibility on their behalf?

RIZOM is therefore not satisfied with just being AI-responsible. It is designed to be agency-preserving and sovereignty-forming.

Whilst responsible AI asks whether the system is behaving acceptably, RIZOM also asks whether the person who uses it remains increasingly able to judge, choose and act.

This reflects the wider purpose of the Fourth Way of AI: not to replace human expertise, but to strengthen the capacity to interpret, orient and act coherently as artificial intelligence becomes embedded in everyday and organisational life.

About the 4th Way of AI

So, What About My Data?

Your personal data should remain under your control. However data sovereignty without interpretive sovereignty is incomplete.

A person can retain formal ownership of their information while gradually surrendering the capacities through which that information acquires meaning. They may control what enters the system yet become dependent on the system to tell them what it signifies.

RIZOM is built to resist that transfer.

• Your words remain yours.

• Your experience is not reduced to a machine-generated identity or label.

• Your symbolic patterns can be made visible without becoming a system’s property.

The technology may accompany the process of interpretation, but it does not appropriate authorship of the meaning produced. The sovereign outcome is not a more comprehensive machine portrait of the individual. It is a person better able to recognise, articulate and act from their own experience.

That is the difference between artificial intelligence that knows more about us and an architecture that helps us remain ourselves.