My illustration entitled: “The Tool That Knows Its Place” — A powerful humanoid AI kneels beside a seated human creator, presenting instruments rather than issuing orders.
Artificial intelligence can be an extraordinary assistant. It must never become an unaccountable authority over the human being it is meant to serve.
Artificial intelligence is moving from the edge of life to its centre. It recommends what we read, filters what we see, predicts what we may buy, helps us write, evaluates applications, detects fraud, prioritises medical attention and increasingly shapes decisions that affect opportunity, reputation and freedom. The question is no longer whether AI will assist human life. The question is whether assistance will quietly become authority.
Human-Sovereignty Transhumanism welcomes technology that expands human capability. It does not treat intelligence, automation or enhancement as enemies. But it insists on a boundary that must not be crossed: no system should acquire final, unchallengeable power over a person’s rights, identity, future or intimate choices.
That boundary is especially important with AI because its power can be deceptively quiet. A machine does not need to issue commands to govern a life. It can govern by ranking, scoring, filtering, nudging, predicting and excluding. When its judgement becomes difficult to understand, impossible to appeal or too widespread to escape, an assistant has become artificial authority.
Technology must remain the servant of humanity. Humanity must never become the property of technology.
The Human-Sovereignty Test offers a practical way to distinguish the two. It asks not only whether a system is accurate, efficient or innovative, but whether it leaves the person more capable of directing their own life. If the answer is unclear, the technology is not yet ready to govern an intimate part of human life.
Assistance is not authority
Assistance expands a person’s ability to decide. It presents relevant information, offers options, explains uncertainty and leaves meaningful room for judgement. An AI writing tool can help a student structure ideas. A diagnostic system can highlight patterns for a clinician. A navigation system can suggest routes. In each case, the tool adds capacity without claiming the final moral or legal decision.
Artificial authority begins when an AI system’s output is treated as decisive merely because it is computational. It appears when a person is denied a loan, job, benefit, service, parole opportunity or medical choice without a real explanation or human review. It appears when automated recommendations are designed to manipulate rather than inform. It appears when a system’s ranking becomes a hidden constitution for everyday life.
The phrase “the algorithm decided” must never be an excuse. An algorithm does not carry moral responsibility. It cannot apologise, repair harm, accept accountability or recognise the dignity of the person affected. The people and institutions that choose the data, objectives, thresholds, deployment conditions and business model remain responsible.
HST therefore does not ask whether AI should be removed from human affairs. It asks whether AI is strengthening agency or replacing it; whether it is assisting human judgement or avoiding human responsibility.
The Human-Sovereignty Test
Before an AI system is placed in a meaningful role, its designers, operators and regulators should be able to answer twelve questions plainly. The test is deliberately human-centred. It examines the real relationship between the system and the person, not the elegance of a technical presentation.
- Ownership: Who owns the data, outputs, profile and practical value created around the person? Does participation transfer unfair control to a platform or third party?
- Control: Can the person understand the system’s role and retain meaningful control over decisions that materially affect them?
- Data: What information is collected, inferred, retained and shared? Is collection necessary, proportionate and protected?
- Voluntariness: Is participation genuinely voluntary, or is consent obtained through economic pressure, default enrolment or lack of alternatives?
- Refusal: Can a person decline AI use without unjust exclusion, punishment or loss of basic services?
- Disconnect: Can the person leave the system, withdraw permissions and recover their information without unreasonable harm?
- Remote alteration: Can a provider change the rules, objectives or operation of the system without meaningful notice and challenge?
- Contestability: Can an affected person ask why a decision occurred, correct relevant errors and obtain independent review?
- Provider failure: What happens if the company fails, is acquired, changes its terms or withdraws access? Does the person lose essential capability, records or standing?
- Accountability: Is there a named human or institution with a legal and practical duty to explain, remedy and prevent harm?
- Equal standing: Does the system unfairly penalise people because of disability, income, age, language, culture, data poverty or a decision not to use enhancement?
- Freedom versus system power: Does the technology increase the user’s effective agency, or does it concentrate power over the user in a distant and opaque system?
These questions must be applied to actual practice, not only policy language. A company may declare that it respects privacy while collecting far more data than it needs. A platform may promise choice while making refusal impractical. A system may offer a nominal appeal while routing every challenge into an automated loop. Sovereignty is not a slogan. It is a lived capacity to understand, choose, refuse, challenge and exit.
Why accuracy alone is not enough
Much of the public discussion around AI focuses on accuracy. Accuracy matters, but it is not enough. A highly accurate system can still be unjust if it is secret, coercive or impossible to contest. A model may identify patterns in historical data while reproducing historical exclusion. It may assign a risk score that is statistically impressive yet morally unacceptable when used to deny a person an opportunity without context.
A person is not a probability. Statistical prediction can assist judgement, but it cannot exhaust the meaning of a human life. People change. Circumstances matter. A person may have reasons, evidence and a story that a model cannot capture. Any institution that uses AI to make consequential decisions must preserve a pathway for those reasons to be heard.
The Human-Sovereignty Test shifts the standard from mere performance to legitimate governance. An AI system should be reliable, but it should also be understandable at the level needed for the decision, auditable by qualified independent parties, and open to meaningful correction. The more serious the impact, the stronger these safeguards must be.
Consent cannot be hidden inside convenience
AI systems often operate through passive collection. People click “agree” because a service is useful, because the alternative is difficult, or because the real consequences are buried in legal language. This is not enough when AI profiles behaviour, infers vulnerability, affects access or trains on information tied to a person’s identity.
Informed consent must be specific and ongoing. A person should know what data is being used, what the system is trying to infer, whether their information will train future models, who receives the output and how long it will be retained. They should be able to change optional permissions without losing unrelated core services.
The standard becomes still higher when AI interacts with the body or mind. Systems involving biometrics, emotion recognition, neural signals, disability support, education or health cannot be treated as ordinary consumer analytics. They operate near the conditions of autonomy itself. Mental privacy, cognitive liberty and bodily sovereignty demand heightened protection.
No employer should quietly require workers to surrender behavioural data to prove attention or productivity. No school should make algorithmic monitoring the price of learning. No public authority should treat a predictive score as a substitute for due process. Convenience is valuable; it is not a licence to convert human beings into permanent sources of extractable data.
My illustration work-in-progress. The art represents the proper balance of power: AI should amplify human creativity and judgment while never claiming authority over the person it serves.
The right to explanation, appeal and human review
A decision that cannot be explained cannot be responsibly imposed. Explanation does not always require revealing proprietary code or every mathematical detail. It requires enough clarity for an affected person to know what happened, what information mattered, how to correct errors and how to challenge the result.
Human review must also be genuine. Placing a person’s name at the end of an automated pipeline is not meaningful control if that person has no time, authority, training or access to depart from the machine’s recommendation. A reviewer must be able to consider context, request additional evidence, disagree with the model and give reasons for the final decision.
This is not inefficient bureaucracy. It is the minimum structure of accountability in a society that remains human. When a decision affects housing, work, health care, education, credit, legal status or freedom, people must not be left pleading with an opaque system that cannot hear them.
AI may calculate. Human institutions must still answer.
Exit rights and resilience against dependency
A person is not sovereign if leaving a system means losing their identity, records, livelihood or access to essential services. This is the same danger that appears in cognitive subscription, now applied more broadly to AI. A platform can make dependence feel like personalisation. It can lock users in through profiles, reputation scores, proprietary histories and networks that cannot be carried elsewhere.
The HST response is practical: people need portability, interoperability and safe exit. Where AI systems manage important personal information, users should be able to obtain usable records, move to another provider and withdraw from unnecessary profiling. Essential services should not be conditional on perpetual consent to commercial surveillance.
Institutions also need resilience. A hospital, school, government office or employer should not place critical human decisions wholly inside a single vendor’s black box. Provider failure, cyberattack, acquisition or abrupt policy change cannot be allowed to erase accountability. Human capability and public oversight must remain available even when a system fails.
Equality in an AI-shaped society
AI can broaden access to knowledge and assistance, but it can also intensify inequality. People with less digital power often have less ability to contest an error, protect their data or choose another service. Communities that are underrepresented in training data may suffer more mistakes while having fewer routes to correction. Those who refuse monitoring or enhancement may be treated as less valuable participants in education and work.
Human-Sovereignty Transhumanism rejects a future divided between those who control intelligent systems and those who are controlled by them. Equal human standing does not mean identical tools or outcomes. It means no person loses their dignity, rights or civic voice because an AI system finds them inconvenient, unprofitable or statistically unusual.
This requires fairness to be more than a technical metric. It requires accessible explanations, non-digital alternatives where essential, anti-discrimination enforcement, independent audits and public participation in the rules that shape collective life. The people most affected by a system must not be the last to know it exists.
Building AI that strengthens human sovereignty
The goal is neither blind acceleration nor fearful prohibition. It is capability with agency. AI should help people learn, create, diagnose, communicate and solve problems while preserving the ability to question, refuse and redirect its use.
For developers, this means designing for user control from the beginning: data minimisation, understandable settings, clear logs, secure architecture, accessible appeal and no dark patterns that manufacture consent. For institutions, it means assigning accountable leaders, testing systems for harm, auditing performance after deployment and preserving human review for consequential decisions. For governments, it means protecting rights that cannot be outsourced to technology: due process, privacy, equality, bodily autonomy and freedom of thought.
For citizens, the central demand is equally clear. We should not accept the false choice between useful AI and human control. We can have both. In fact, control is what makes a technology worthy of trust.
The final authority remains human
Artificial intelligence may become more capable, more present and more persuasive. It may assist millions of decisions each day. But intelligence is not authority, and computational power is not moral legitimacy. The final authority over a human life must not migrate into a model, a platform or a dashboard.
The Human-Sovereignty Test keeps the first question where it belongs: does this system serve the person, or does the person become a function of the system? If an AI cannot pass that test, its deployment must be limited, redesigned or refused.
Human-Sovereignty Transhumanism is a commitment to progress without surrender. Let artificial intelligence assist humanity. Let it expand our capabilities. But let it never inherit artificial authority over the human being.