Artificial Intelligence and Freedom of Knowledge: Who Decides What People Are Allowed to Know?

Knowledge is widely regarded as one of the most important foundations of a free society. People must be able to ask questions, seek information, compare different sources, and form their own opinions. Yet this freedom increasingly depends on which information digital platforms, search engines, academic databases, and artificial intelligence systems make accessible.

Today, knowledge is rarely prohibited openly. Restrictions usually operate in far more subtle ways. Certain content becomes harder to find, sources do not appear in search results, questions are rejected by artificial intelligence systems, and answers are shortened or steered in an institutionally preferred direction. In other cases, an answer appears complete even though important information or alternative perspectives are missing. Knowledge may still exist in theory, but in practice it becomes accessible only when a technical or institutional system permits access.

This raises a fundamental social question: Should governments, corporations, universities, platforms, or artificial intelligence providers be allowed to decide what an adult may know, investigate, or understand?

The answer must be clear. No institution possesses a general right to prohibit knowledge as such or to restrict access to lawful information merely because that information could potentially be misused. Concrete harmful conduct may be regulated. Knowledge, research, understanding, and personal insight, however, must not be equated with harmful action.

Knowledge Is Not Conduct

The most important distinction in this debate is the separation between knowledge and conduct. A person may learn about a subject without intending to apply what they have learned. They may examine a controversial position without adopting it. They may want to understand how a dangerous act works without carrying it out. Likewise, a scientist may investigate a troubling phenomenon without becoming part of that phenomenon.

Several stages separate information from an actual act. Information is first received and evaluated. It may then be accepted, rejected, or examined further. Only later could it lead to intent, preparation, and eventually a concrete action. These stages must not be confused with one another, either legally or socially.

A question is not an intention. Reading information is not preparation. Understanding is not execution. Research is not proof of criminal purpose. The mere possession of knowledge is not itself a harmful act.

Laws may restrict violence, fraud, extortion, violations of privacy, conspiracy, concrete preparations for a crime, and direct participation in harmful conduct. This does not create a general right to deny adults access to knowledge. The legal boundary must be drawn where concrete action begins, not at the level of thinking, questioning, or understanding.

How Modern Knowledge Control Works

Earlier forms of censorship were often visible. Books were banned, printed materials were confiscated, or certain opinions were openly suppressed. Modern knowledge control, by contrast, often operates through access. Information is not necessarily destroyed. Instead, it is categorized, filtered, or hidden in ways that make it difficult for ordinary users to find.

Search engines determine which sources become visible through their ranking systems. Academic databases determine which publications can be found. Platforms may restrict content or remove it from recommendation systems. Universities and funding institutions influence which topics are researched through access requirements, funding decisions, and evaluation systems. Artificial intelligence can reject questions, reframe answers, omit certain sources, or present only a narrow portion of a subject.

This can be described as censorship at the access layer. The information still exists, but the practical route to it is obstructed. This kind of restriction is especially difficult for users to detect. They see only what the system displays. What is missing may appear not to exist at all.

This can create a situation in which knowledge is formally free, while actual access depends on institutional permission.

Artificial Intelligence as an Intermediary of Knowledge

Artificial intelligence is increasingly taking over tasks that were once distributed across multiple systems. It searches for information, summarizes sources, explains scientific relationships, translates technical language, answers legal questions, and assists in developing new research questions. For many people, it is becoming the primary interface between a question and the available body of knowledge.

This role gives AI systems considerable influence. They do not merely determine which answer is given. They also influence which sources are considered, which perspectives appear relevant, and which connections are shown to the user at all.

A system may reject a question directly. It may shift to another topic, insert moral judgments, or provide only general information. At least this kind of refusal is visible. More problematic is an answer that sounds detailed and convincing while systematically omitting essential information.

This can be described as synthetic censorship. It occurs when a system produces an apparently complete answer while excluding relevant and lawful information because of internal rules, institutional interests, or political requirements. The user may not realize that only a limited portion of the available information has been presented and may confuse the limits of the system with the limits of knowledge itself.

The confident language of modern AI systems intensifies this problem. A fluently written answer often appears credible even when it is incomplete, one-sided, or supported by unverifiable sources. It is therefore not enough for an artificial intelligence system to answer politely and clearly. It must also disclose when it does not know certain information, cannot retrieve it, or is withholding it because of particular rules.

Unequal Access to Knowledge Creates Unequal Power

Not everyone has the same access to information. Governments, intelligence agencies, major corporations, military institutions, and well-funded universities often possess powerful models, internal databases, extensive archives, specialized experts, and substantial computing capacity. The general public, by contrast, usually receives simplified systems, limited search interfaces, paywalled databases, and filtered summaries.

This creates a structural inequality of knowledge. One group can examine information in detail, compare multiple sources, and act on a broad base of knowledge. Another receives only a limited selection and must rely on the claims of institutions that themselves possess far greater access.

This inequality is not merely a matter of education or convenience. It affects the ability to verify political claims, understand economic decisions, defend legal interests, and identify institutional abuses of power.

It is especially contradictory when adults continue to be held responsible for their decisions while being denied access to the information necessary for independent judgment. A society cannot treat people as legally and morally responsible adults while simultaneously claiming that they are not mature enough to know certain things.

The Power of Private Companies Over Access to Knowledge

A company generally has the right to decide which products and services it offers. This freedom becomes problematic, however, when a small number of major providers control the most important pathways to knowledge. At that point, the issue is no longer merely an ordinary product decision. It becomes the management of social infrastructure.

Search engines, AI systems, cloud services, academic databases, social networks, and educational platforms are now closely interconnected. Whoever controls these systems can exert considerable influence over which information is discoverable, credible, and socially acceptable.

It is not enough to point out that alternative providers theoretically exist. A real alternative exists only when switching is practical. The other system must offer comparable capabilities, languages, sources, technical functions, and affordable access. A weak or barely usable alternative does not create genuine freedom of choice.

When users cannot move to another system without significant disadvantage, a form of structural dependence develops. The provider can restrict access to knowledge without those affected having a realistic way to leave.

Safety and the Possible Misuse of Knowledge

Many restrictions on knowledge are justified in the name of safety. Certain information may be dangerous or may be used for harmful purposes. This is a real issue because scientific and technical knowledge can often support both beneficial and harmful applications.

Knowledge of chemistry can be used to develop medicines or poisons. Computer science can protect systems or attack them. Biology can help treat disease or explain dangerous processes. Psychological knowledge can support people or be used manipulatively.

The possibility of misuse, however, does not mean that knowledge itself may be prohibited. Otherwise, nearly every form of education could be restricted. The decisive question is whether information is being explained in general terms or whether someone is directly participating in a concrete harmful act.

A scientific overview, a historical account, a theoretical explanation, or a comparative analysis remains knowledge. A different assessment would apply to individualized support tailored to a specific target and directly contributing to the execution of a harmful act.

Controlled materials, illegal procurement, concrete preparatory actions, and direct operational participation may therefore be regulated. The general communication of knowledge must not be equated with these acts.

How Research Can Be Restricted at the Level of the Question

Knowledge control does not affect only completed answers. It can also determine which questions are asked and scientifically investigated in the first place.

Research depends on funding, universities, ethics committees, academic journals, databases, and technical platforms. When certain topics are regarded as undesirable, politically problematic, or institutionally inappropriate, they may not be funded, published, or made fully accessible.

This can create a cycle. A topic is initially studied very little. Later, the small number of studies is presented as evidence that insufficient knowledge exists. This alleged lack of evidence is then used to justify further reluctance to support research or publication.

The absence of studies does not always mean that a subject lacks scientific importance. It may also mean that research was prevented, obstructed, or never funded.

Artificial intelligence can intensify this effect. When it does not search certain sources, rejects particular questions, or presents only institutionally preferred perspectives, it shapes the starting point of scientific work. Researchers may fail to develop certain hypotheses because relevant sources or alternative approaches remain invisible to them.

Scientific freedom must therefore begin with the question itself. It includes not only the right to publish findings, but also the right to test unconventional hypotheses and challenge existing assumptions.

Government Control Through Private Platforms

Another problem arises when government agencies do not openly exercise information control themselves but instead pressure private companies to do so. This can occur through laws, liability risks, licensing requirements, public contracts, informal discussions, or political pressure.

The restriction then appears to be the voluntary decision of a company, even though it was actually shaped by government interests. In this way, a state can use private platforms to implement measures that would be subject to legal review if imposed openly by the government.

What a state may not directly and lawfully prohibit should not be enforced indirectly and without transparency through private providers. When government influence and private infrastructure jointly determine access to knowledge, clear legal rules are necessary. Restrictions require a legal basis, a reasoned explanation, an assessment of proportionality, and a meaningful opportunity to appeal.

Secret arrangements or informal pressure must not determine which lawful information adults receive.

Knowledge, Education, and the Interpretation of History

Artificial intelligence does not merely communicate isolated facts. It explains causes, places historical events in context, and connects information into a coherent narrative. In doing so, it determines which events are emphasized, which sources are used, and which perspectives are included.

History, however, does not consist only of data. It includes different experiences, interpretations, and conflicts. When a small number of AI systems turn this diversity into a single and apparently neutral account, a dominant standard narrative can emerge.

This development is especially significant in education. Students may become accustomed to accepting a finished answer instead of reading primary sources, comparing conflicting positions, and evaluating uncertainty for themselves.

The solution cannot be to prescribe one officially correct or balanced perspective. That would merely replace one form of control with another. What is needed are different systems, traceable sources, access to original documents, and the ability to compare multiple accounts.

Education should not merely provide people with information. It should enable them to examine, distinguish, and judge independently.

When People Move Into Alternative Knowledge Spaces

People do not stop seeking knowledge simply because major platforms reject certain questions. They move to foreign systems, local AI models, anonymous networks, decentralized archives, private groups, or lesser-known platforms.

This movement is not always voluntary. It may occur because established systems no longer provide sufficient access. In that case, it amounts to forced migration into alternative knowledge spaces.

This can create new problems. Alternative sources may be harder to verify, technically insecure, or more vulnerable to manipulation. Users may face greater risks of fraud, malware, or surveillance.

Restrictions on knowledge can therefore intensify the very dangers they claim to prevent. When controversial topics are pushed out of open and verifiable environments, discussion moves into closed spaces where errors and misinformation are more difficult to correct. Open access with verifiable sources is therefore often safer than a policy of concealment.

Requirements for a Free Knowledge Order

Freedom of knowledge must not exist only as an abstract right. It requires practical conditions. Systems must disclose whether they cannot answer a question, do not know the information, or are refusing to provide it because of a particular rule. Silent omission must not be presented as a complete answer. Sources must be discoverable and verifiable. A citation is worthless if the referenced document does not exist or cannot be accessed by the user.

Users must also have the ability to challenge decisions. A restriction should be explained, reviewable, and correctable. A diversity of providers is equally important. No single company or institution should have final authority over which questions adults may ask and which information they may receive. Data portability, open formats, local systems, and independent archives can reduce this dependence.

Government influence must also remain transparent. Public authorities should disclose when they ask platforms or AI providers to restrict lawful information.

In research and education, it is necessary to document the model versions, databases, search conditions, and relevant restrictions that were used. Only then can it be determined whether an answer genuinely reflects the available body of knowledge or merely the output of a limited system.

Knowledge Must Not Become a Permission-Based Activity

The natural limit of knowledge lies in a person’s present capacity to understand it. That capacity should be expanded through education, research, and experience. It should not be artificially limited through institutional prohibitions or invisible filtering systems.

Institutions may explain, contextualize, criticize, and identify uncertainty. They may help people evaluate sources and assess the quality of evidence. Their role, however, must not be to control adults through ignorance.

A population that may know only what an institution considers appropriate is not truly autonomous. It remains dependent on those who control access to information. No government, corporation, university, platform, or artificial intelligence provider possesses a general right to determine what adults are allowed to know. Knowledge may be discussed, examined, criticized, and disproved. Concrete harmful conduct may be prohibited and punished. The boundary must remain clear: knowledge belongs to free inquiry, while responsibility and liability belong to concrete action.

The complete academic article has been published in the Advanced International Journal of Multidisciplinary Research:

Elias Rubenstein: When Knowledge Requires Permission: Artificial Intelligence, Epistemic Asymmetry, and Institutional Power
DOI: in process

Elias Rubenstein: When Knowledge Requires Permission: Artificial Intelligence, Epistemic Asymmetry, and Institutional Power.pdf