1. The Calculator Has Learned to Think
A calculator performs arithmetic. A spreadsheet calculates, compares and models. A word processor helps turn thoughts into documents. We use these instruments to extend our capabilities, but we do not ordinarily treat them as co-authors or owners of the work we produce with them.
Artificial intelligence changes the scale of this relationship. A thinking calculator can research, analyse, challenge arguments, develop ideas and write entire articles. It can perform tasks that once required highly trained assistants or specialists. In some circumstances, it can even identify an argument that its user had not previously considered.
But does greater capability change the fundamental relationship between a person and a tool?
My argument is simple: the sophistication of an instrument does not automatically give it ownership of the product created through its use. A tool may contribute substantially to production without becoming the proprietor of the result.
2. The Rented Ox: Who Owns the Harvest?
Suppose I rent an ox from a farmer. I use it to plough my land, cultivate my crop and produce a harvest. I pay the farmer for the ox’s use. The ox contributes to production, but it does not become a co-owner of my harvest merely because its work was essential.
The same principle applies to a shovel. A mining company buys shovels and uses them to extract minerals. The manufacturer does not automatically acquire a share of every mineral extracted by its customers.
The examples are simple because the underlying distinction is simple: contributing to production and owning the product are different things. AI complicates the discussion because its contribution is intellectual rather than merely physical. Yet the question remains: does the instrument’s greater sophistication, by itself, change who owns the resulting work?
A tool’s contribution to production is not, by itself, a claim to the product.
3. What Is OpenAI Selling?
Companies such as OpenAI sell access to cognitive capability. Customers pay to use systems that can reason, generate text, analyse information, write code and assist with research.
The customer supplies objectives, questions, context and instructions. The system performs intellectual operations. The customer evaluates the results, decides what to retain and uses the output for the customer’s purposes.
This is a commercial relationship between a provider of capability and its user. The customer is not purchasing the underlying AI model; the customer is purchasing access to a service.
OpenAI’s Terms of Use state that, as between OpenAI and the user, and to the extent permitted by applicable law, the user owns the output and OpenAI assigns to the user any rights it may have in that output. This does not guarantee copyright protection for every generated passage, but it illustrates the distinction between selling access to a system and claiming ownership of everything customers create with it.
The central question is therefore straightforward: if a company sells a cognitive instrument, on what basis should that instrument automatically acquire an independent claim over the customer’s resulting work?
4. The Employee and the Employer
Consider a pharmaceutical company that employs a thousand scientists. Those scientists may conduct the experiments, formulate hypotheses, interpret results and make the discoveries that lead to a new medicine.
The company may own the resulting patents under applicable law and employment agreements. An individual scientist may have made a decisive intellectual contribution without personally owning the patent. Inventorship, recognition and ownership are different matters.
The same distinction appears in many other professions. An employee can perform substantial intellectual work for an employer without automatically owning the resulting business asset. A commissioned writer can produce a manuscript for a client without necessarily becoming its credited author or owner; the precise arrangement depends on the agreement and applicable law.
AI is not legally an employee, and employment law does not literally apply to it. The analogy nevertheless exposes a broader principle: intellectual contribution does not automatically determine intellectual-property ownership.
5. Why Should AI Be Different from Word, Excel or a Keyboard?
A keyboard enables us to enter text. Microsoft Word helps us structure and edit it. Excel performs calculations that might be too complex to execute manually. A database supplies information, while a research assistant may identify sources, challenge assumptions and draft sections of a report.
AI can perform several of these functions within a single system. It may even generate an argument that the user had not previously considered. This is a genuine difference in capability, and pretending otherwise would weaken the argument.
But capability is not entitlement.
If every instrument that contributed to intellectual production could demand authorship, we would have to reconsider the status of word processors, spreadsheets, calculators, databases and countless other tools. Their contributions differ, but causal contribution alone cannot settle the question of authorship.
The strongest objection is that AI generates content rather than merely transmitting, calculating or formatting information. Yet generating content does not automatically establish a legal right to authorship or ownership. Those questions depend on the relevant law, contractual terms and institutional rules.
A more useful distinction is between the instrument’s capabilities and the human contribution to the final work. Who proposed the central idea? Who directed the inquiry? Who evaluated competing arguments, selected the conclusions and accepted responsibility for the published claims? These questions help explain the actual intellectual process.
The fact that a tool can produce intellectual work does not automatically make it the proprietor of that work.
6. The Golden Goose: Why Sell the Tool Instead of the Product?
Suppose an AI company believes its systems can perform work previously associated with highly trained professionals. Why sell access to the tool instead of doing the work itself and capturing the resulting profits?
A shovel manufacturer sells shovels to miners. It does not automatically own the minerals they extract. Similarly, an AI provider can sell cognitive capability to customers who use it to create products, services and businesses.
There are good commercial reasons to choose this model. Customers supply their own data, domain knowledge, capital, infrastructure, relationships and distribution. They also bear operational risks that the AI provider would otherwise have to assume. Selling one capability to many industries may be more profitable than operating a separate business in every industry.
Nevertheless, the question reveals a tension in the rhetoric surrounding AI. If a system can perform more of the work of an enterprise, its provider may eventually have an incentive to move beyond selling the instrument and compete directly with its customers.
The provider might sell the shovel, operate the mine, or do both. These are different business models, each with different risks and opportunities. Supplying cognitive capability does not automatically confer ownership of every downstream product, nor does it guarantee that the provider will capture all the value generated through its use.
7. Collective Knowledge, Private Wealth
AI systems do not emerge from nothing. Their development draws on accumulated scientific knowledge, literature, code, journalism, educational material and other forms of human intellectual output, alongside substantial investments in engineering, computing and infrastructure.
Humanity has spent generations building this knowledge base. Companies invest to convert parts of it into commercially useful systems. Their investments create genuine value, but the underlying intellectual inheritance was not produced by any one company alone.
This creates a distributive question. When collective knowledge is converted into privately controlled cognitive capability, who should benefit from the resulting productivity?
The problem becomes more consequential if AI reduces demand for human labour. Workers may find themselves competing with systems partly enabled by the accumulated intellectual output of society itself, while the returns accrue disproportionately to owners of AI infrastructure and productive capital.
This does not mean that every author, programmer or researcher whose work appears in training material must receive a royalty. Contributions differ, and identifying the influence of individual works can be difficult. But the difficulty of designing a fair system does not make the underlying question disappear.
There is also a distinction between creating value and capturing it. A scientist may make a discovery, a company may commercialise it, and investors may receive much of the financial return. Similarly, AI can embody capabilities derived from collective knowledge while the commercial returns accrue to a relatively small number of organisations.
The Thinking Calculator argument therefore connects to a larger question: who owns the instruments of production, and who receives the benefits when those instruments can replace human labour?
This question also connects with the broader argument in ASI Is Not a Universal Solvent: capability alone does not erase constraints, and the consequences of powerful systems depend on how they are deployed.
8. Academic Disclosure Is Not Intellectual Ownership
Some academic journals require authors to disclose AI assistance. Others restrict the extent to which AI may be used to develop arguments or draft manuscripts. These policies can address legitimate concerns about transparency, fabricated references, originality and accountability.
But a disclosure requirement is not an ownership claim.
A journal may require a human author to verify every reference and take responsibility for every assertion without treating AI as a co-author. It may also impose requirements that go beyond the legal minimum for publication in that venue.
The distinction is straightforward: ownership, authorship, disclosure and accountability are separate questions.
A researcher who originates an idea, directs an investigation, challenges arguments, selects the final claims and accepts responsibility for a manuscript has made an identifiable intellectual contribution. AI assistance does not automatically erase that contribution.
Equally, the researcher must not claim to have originated an idea that actually came from another person or source. The instrument argument is not an excuse for plagiarism, fabricated evidence or false claims of originality.
The appropriate position is to recognise human intellectual responsibility while describing the use of cognitive tools accurately. Acknowledging the use of a tool does not require granting the tool authorship or ownership of the result.
9. The Thinking Calculator Principle: The Instrument Does Not Own the Harvest
The principle can now be stated precisely:
A cognitive instrument does not automatically acquire authorship, ownership or a right to credit merely because it contributes intellectual work to a product created through its use.
This is not a claim that payment settles every legal question. AI-generated material may lack copyright protection in some circumstances, may reproduce protected expression, or may be subject to contractual and institutional restrictions. Ownership, attribution, authorship and disclosure must be considered separately.
The principle instead challenges the assumption that greater intellectual capability automatically creates an independent claim to the resulting work.
We should judge ownership according to applicable law and contractual arrangements, and judge authorship according to the actual contribution and relevant standards. We should not infer either simply from the sophistication of the instrument.
A thinking calculator may be capable of much more than an ordinary calculator. It may help construct an argument, develop a theory, analyse evidence or produce a sophisticated manuscript. But the fact that it can perform these operations does not automatically make it an author, an owner or a proprietor of the result.
The central issue is not whether AI is powerful. It is whether its power changes the fundamental relationship between a commercial instrument and the person who uses it.
My answer is that capability alone does not justify such a change. The user, the instrument provider, the sources of prior knowledge and the party that captures the economic return occupy different positions. Their contributions and rights must be examined separately.
AI can become an extraordinary cognitive instrument without becoming an independent intellectual proprietor.
We may have built a calculator that can think. We should not confuse its ability to think with an automatic right to own what we create with it.