My illustration entitled: “The Staircase Beside the Lift” – A spectacular mechanical lift rises through a monumental tower, with an elegant, accessible staircase winding alongside it.
Imagine spending months developing a project with an AI assistant. It helps organise your notes, compare ideas, prepare drafts and remember why earlier decisions were made. The work moves faster, and you can attempt things that would otherwise have been difficult.
Then the service changes its conditions. You can accept them or leave.
Leaving sounds straightforward until you discover what it means. Your notes can be downloaded, but their connections are lost. Your drafts can be read, but important parts cannot be edited elsewhere. The reasoning behind earlier choices remains inside conversations you cannot easily recover.
You still have a cancellation button. Your practical ability to continue has become uncertain.
This is the concern behind the right to remain capable. I use the phrase to express a moral principle: people should be able to benefit from intelligent assistance while retaining a usable capacity to understand, direct and continue their own activities.
For Unceded Intelligence, revocable authority requires more than a setting that withdraws permission. As reliance grows, people need judgment, appropriate skills, accessible alternatives and continuing access to their work.
The freedom to stop using a system matters only if stopping leaves a practicable way to carry on.
Capability is more than a completed task
An assistant can produce an impressive result while leaving its user poorly equipped to evaluate, revise or reproduce what matters about it.
This does not make the result worthless. It does mean that output alone is an incomplete measure of empowerment.
Consider a person receiving a detailed project plan. The plan may be useful, but can they identify its assumptions? Can they change its priorities? Do they know which information needs checking? Can they continue the work if the original assistant becomes unavailable?
A completed task and a retained capability are connected, but different.
Capability includes the practical means to act: understanding what one is trying to achieve, recognising relevant choices, accessing necessary materials and obtaining help on acceptable terms.
Some of that capability can reside in tools and relationships. It need not all exist inside an individual’s memory. The question is whether the arrangement preserves a person’s effective ability to direct their affairs, including when a particular provider changes or disappears.
Judgment should become better informed
AI assistance can help people consider possibilities they might otherwise miss. It can organise competing arguments, explain unfamiliar terms and reveal questions worth asking.
These benefits are strongest when they support judgment.
A person should be able to examine why a recommendation fits their purpose, what assumptions it makes and what significant alternatives remain. Confidence of presentation should not stand in for reasons.
Meaningful understanding does not require tracing every internal calculation. It requires enough grasp of the decision to recognise what is at stake and seek appropriate verification.
That standard varies with the activity. Choosing wording for an informal invitation is different from approving a consequential commitment. The degree of scrutiny should reflect the consequences of error.
A helpful system can offer explanations, identify uncertainty and make disagreement easier. It should also leave room for a person to revise the objective itself.
Someone may understand the recommendation perfectly and still decide that a different consideration matters more.
Retained judgment means that the person can question the destination as well as inspect the proposed route.
Which skills should remain available?
Delegation changes what people practise. When a tool takes over an activity, the user may have fewer occasions to exercise the corresponding skill.
That possibility deserves attention, but it does not establish an obligation to preserve every previous way of working.
Much useful technology allows people to stop doing burdensome tasks. Requiring everyone to maintain every manual skill would undermine the point of assistance and exclude people who benefit most from it.
The better question is which abilities matter to the person’s purposes, responsibilities and ability to recover from failure.
A writer may value retaining the ability to develop an argument and revise a paragraph. A project manager may need to recognise an unrealistic schedule. Someone supervising a consequential process may need enough knowledge to identify when an apparently reasonable answer requires review.
The necessary skill may be checking, interpreting or asking for help rather than performing the entire task unaided.
Systems can support this through optional explanations, visible intermediate work, opportunities to practise and assistance that can be reduced when the user wants greater involvement.
Education and workplaces also have responsibilities here. They should consider which competencies people need to develop before expecting them to supervise automated work.
The aim is to preserve useful agency while allowing the content of human skill to evolve.
Dependence can increase independence
Practical independence should not be confused with complete self-sufficiency.
Human beings depend on language, infrastructure, teachers, colleagues, institutions and devices. A person who uses assistive technology may gain substantial freedom through reliance on that technology.
Asking whether they can perform the same activity without any assistance would misunderstand the achievement.
The relevant concern is whether reliance leaves them vulnerable to arbitrary control by a particular provider.
A person may need a communication aid while still deserving choices about its operation, access to their records and protection when changing services. Their reliance strengthens the case for continuity; it does not weaken their standing as someone entitled to direct their life.
Independence can be maintained through multiple providers, trusted human support, shared infrastructure and enforceable institutional obligations. An alternative does not have to mean doing everything alone.
The objective is to prevent reliance from becoming a claim of ownership over the person who relies.
Access to your work must remain useful
A person should be able to retain usable access to the materials they contribute and the work they are entitled to take with them.
The practical meaning of access depends on what the work requires.
A picture of a document may preserve its appearance while losing the ability to revise its structure. A list of notes may preserve individual sentences while losing the links that made them intelligible. A project archive may contain the final result without the instructions, sources or decisions needed to continue it.
Portability should therefore be judged by whether the person can resume the relevant activity elsewhere.
Where feasible, that requires editable formats, intelligible organisation, necessary attachments and enough context to understand how the work developed.
People should also be able to distinguish their contributions, the system’s suggestions and statements that still need verification.
This principle must respect other people’s confidentiality and legitimate restrictions on shared material. Access to one’s work does not create unlimited entitlement to every dataset, private record or proprietary component used by a service.
But those limits should be explained precisely. They should not become a broad excuse to make a person’s own recoverable contributions inaccessible.
My illustration “The Staircase Beside the Lift” work-in-progress. The art represents that people should be able to freely choose either route: convenience preserves the means to proceed independently.
An alternative must be usable by the person concerned
The existence of another product does not automatically establish meaningful choice.
An alternative may be unaffordable, inaccessible, incompatible with necessary materials or unsuitable for the language in which a person works. Moving may require time or expertise that the person cannot reasonably obtain.
These differences matter when evaluating whether withdrawal is practical.
A replacement need not reproduce every convenience of the original service. Some disruption and loss of preferred features may be unavoidable. The important question is whether essential activities can continue on reasonable terms.
For one person, that might mean using another application. For another, it might mean a simpler process supported by a colleague or community organisation.
Providers and institutions can improve those options through interoperable formats, transfer assistance and clear documentation. They can also avoid designing unnecessary barriers that make their own service artificially difficult to replace.
Freedom of choice is strengthened when alternatives are ordinary arrangements people can actually use.
Plan for withdrawal before dependence deepens
The best time to consider exit is before a service becomes the sole place where an important activity can happen.
That does not mean treating every useful relationship with suspicion. It means taking continuity seriously as part of responsible design.
A service should make clear what happens when someone pauses assistance, withdraws a particular permission or closes an account. Those actions should remain distinguishable.
A person might want to stop automated replies while retaining access to correspondence. They might reject a new use of their information while continuing to use an established function. They might need time to transfer a project before ending the relationship.
Where feasible, reducing permissions should preserve the functions that do not depend on them.
Transition arrangements should address outstanding actions, recurring instructions, connected services and access to necessary records. If a safe transition takes time, that need should be explained and bounded.
A provider cannot always promise indefinite support for an old system. It can, however, communicate changes early, offer reasonable recovery opportunities and avoid making departure needlessly destructive.
The cost of saying no
Withdrawal may carry real costs. Transferring a complex project can take effort. A replacement may work differently. Some services depend on infrastructure that cannot be reproduced without expense.
Recognising those costs is compatible with opposing penalties designed to suppress refusal.
There is a meaningful difference between the effort inherent in moving and an artificial obstacle imposed to prevent movement.
Consider a provider that allows an ordinary export while explaining its limitations. Compare it with one that withholds recoverable work unless the user accepts an unrelated expansion of data use.
The latter arrangement turns accumulated reliance into pressure to surrender additional authority.
The more central a service becomes to someone’s work or everyday participation, the more seriously its provider should take the consequences of refusal and withdrawal.
Usefulness creates responsibilities. It should not create an unrestricted entitlement to dictate new terms.
Institutions must preserve routes to participation
Practical independence is not solely a matter of individual preparation.
An employee may have little influence over the systems an employer requires. A student may be unable to choose the platform through which essential teaching is delivered. A customer may encounter an automated process because an institution has made it the only route to service.
In those circumstances, telling the person to switch providers or maintain a private backup may fail to address the source of dependence.
The institution should examine how people can access their work, correct mistakes, obtain appropriate human assistance and continue essential activities when a system fails.
It should also distinguish functions that genuinely require automation from optional uses that can be declined without excluding someone from participation.
Where individual choice is constrained, accountability and continuity become more important.
The burden of preserving freedom should not fall entirely on the person with the least power to shape the arrangement.
Testing whether independence remains practical
A useful assessment begins with concrete questions:
- Judgment: Can the person understand the important choices, question recommendations and seek an independent view?
- Skills: Do they retain or have access to the abilities needed to supervise, revise or continue the activity?
- Work: Can they recover the materials and context necessary to carry on elsewhere?
- Alternatives: Is there a usable replacement, supported process or fallback for what matters most?
- Withdrawal: Can they narrow or end permission without avoidable loss or disproportionate pressure?
- Responsibility: Who must help when dependence prevents a person from exercising those choices?
The answers will vary across people and activities. A system that preserves independence for an expert may leave an ordinary user unable to move. Accessibility, resources and institutional position belong in the assessment.
These questions are intended to guide improvement. They should not become another demand that vulnerable users prove their self-sufficiency before receiving assistance.
Intelligence that leaves people able to choose
The promise of intelligent assistance is larger than making people faster at completing tasks. It includes helping them understand more, attempt more and participate more fully in activities they value.
That promise is diminished when the same assistance makes their work inaccessible, their judgment harder to exercise or their alternatives impossible to use.
Unceded Intelligence asks us to preserve capability throughout the relationship: when assistance begins, when its role expands and when someone decides to change or end it.
I want systems that return time without taking away the means to act. I want assistance that can become deeply useful while remaining open to challenge, revision and replacement.
The right to remain capable means preserving a person’s practical standing as the director of their activities, including through the tools and people on whom they choose to rely.
I may rely on intelligence without surrendering my capacity to judge, my access to my work or my ability to continue on different terms. The freedom to delegate remains secure only while the means to redirect, replace or leave survive.