Who Owns the Idea?

In Sumer, the gods held the Me.
In Egypt, Thoth wrote the words before you did.
In Athens, you asked a Muse.
In 2026, you open a tab and ask a model.

Working all day with advanced AI systems is fascinating. There is a lot of coding, but for a communicator like me, it’s interesting and oh-so-satisfying to see your ideas take shape. The fascinating bit comes from the combination of structure, psychology, philosophy, and pure creativity. And then sometimes a particular idea or a solution comes from across the ether onto your screen, and the idea is original and perfect and purely from an AI, which pushes larger inquiries. If I didn’t generate it, do I own it? Am I accountable for it?

For most of recorded history, the question “where do ideas come from?” was a spiritual one, with an answer rooted in culture and tradition that said something important about who you were and what you were responsible for.

Skip to today, and it seems we’ve swapped that ancient question for efficiency. Most organizations haven’t noticed what they gave away in the trade.

The oldest written framework for where ideas come from doesn’t belong to the Greeks (yes, if you read some of my other posts, you would be justified in thinking that I was going there, and I will, but not quite yet). It belongs to the Sumerians, a civilization in southern Iraq that was already ancient when Athens was a village.

Their concept was the Me (pronounced “may”): a set of divine powers that underpinned all of civilization — writing, judgment, music, wisdom, craftsmanship, law, and even descent and lamentation. The Me existed before any human held them. In the myth of The Gifts of Inanna, the god Enki, patron of wisdom and the sciences, holds the Me and transmits them to Inanna, who carries them to humanity. Ideas, in this framework, were not generated. They were received, transmitted, and stewarded. The human was the channel. The Me was the source.

Make sure to read the sobering fine print. The Me also included darker forces, deceit, envy, and warfare. The Sumerians understood that the same divine transmission that gives you wisdom also gives you the capacity for catastrophic misuse. The gift came with a shadow attached. Ancient Mesopotamian scribes performed a ritual acknowledgment before beginning their work: the knowledge they used had a source, and they were accountable to it.

Maybe you get where I am going with this, but not just yet, since the journey in this case is so much more interesting than the destination. So now, Egypt-bound, where and when knowledge was morally weighted.

In Egypt, hieroglyphics that we have all seen in museums had a name: medu netjer, “the words of the gods.” Writing was not invented by humans. It was given by Thoth, god of wisdom, knowledge, and cosmic balance, a deity venerated continuously for thousands of years, from the earliest dynasties through the Ptolemaic period. Before starting work each day, scribes would pour out a drop of ink in Thoth’s honor. The tradition held that Thoth had encoded all knowledge humanity would need into 42 books, some reserved as sacred, revealed only to initiates.

But the more structurally interesting element is Maat, not just truth or justice, but the order of the cosmos itself. Ideas aligned with Maat carried cosmic weight. Ideas that violated it were existentially wrong, not just incorrect. At death, your heart was weighed against Maat’s feather. This is a system where knowledge is always already morally loaded, not neutral, not a resource. Every idea carried accountability built in.

Nope, we haven’t arrived at our destination yet. Board the next ship further with the light of the Pharos of Alexandria in our eyes, and off to West Africa and then to the Australian desert, where the self you choose, the land that remembers becomes beauty in truth.

In Yoruba cosmology, Orunmila is the deity of wisdom and omniscience, the Witness of Fate, present at the creation of all souls. But the more striking concept is Ori: the personal divine consciousness that each soul chooses before birth, the inner source that shapes all decisions, creativity, and thought throughout a life.

This is close to the Roman genius, but with a crucial difference. You don’t receive Ori from outside. You select it before incarnating. Authorship and responsibility are entangled from the start, ideas don’t come from a god external to you, they come from the Ori you chose, which is inseparable from who you are. The Ifa divination system that Orunmila governs functions as a vast knowledge framework: wisdom is already encoded in the corpus; the human task is to access it correctly and interpret it for the specific situation. A system that holds the knowledge. A person who is responsible for the application. Sound familiar? I know that it was at this point in my research that I was nodding at the screen, thinking, “Wow, AI is not that unique in the way it builds.”

A bit more patience, and we are getting there.

The most radical version of this framework belongs to Aboriginal Australians, the oldest continuous culture on earth, spanning more than 50,000 years. In their tradition, the Dreaming (Jukurrpa) is not a creation story confined to the past. It is a permanent reality that exists across the past, present, and future simultaneously. Ancestor Beings encoded knowledge, law, and meaning into the land itself. Songs, stories, ceremony, and art are not representations of ideas. They are the ideas, maintained in the country.

In this framework, ideas don’t come from individuals at all. They come from the land, the Ancestors, and the relational practice that keeps the transmission alive. The human is custodian and transmitter — not creator. To transmit poorly, or to transmit sacred knowledge to the wrong context, is a breach with consequences that extend across community and country. Accountability more total than anything most organizations have ever designed.

Ok, so finally we are here, you are thinking, and you would be right. It’s a bit long-winded, but often the best philosophical points are, and who am I to muddy tradition?

So now we fast-forward across fifty centuries. When you type into an LLM, you are not invoking Thoth or asking the Muses or consulting Orunmila. You are tapping something closer to an industrial commons, a system trained on the aggregate of written human thought, optimized to return the most statistically probable next word given what you asked. It is, by design, an averaging engine.

The research is becoming hard to ignore. A 2024 study of 300 writers found that AI assistance improved individual outputs but significantly reduced collective diversity, pulling everyone’s work toward the same center. A 2025 paper in Nature confirmed that AI models trained on AI-generated content undergo “model collapse”: successive generations lose the edge cases, the rare inputs, the things that make outputs genuinely distinct, eventually degrading into what researchers described as “unrelated nonsense.” The World Economic Forum noted plainly that models trained on the same data will, over time, generate the same outputs. And I and others have published numerous posts and articles on this right here in this forum.

Yet, research published in Nature Human Behavior in early 2026 went further, showing that AI adoption is producing a scientific monoculture converging not just in what is studied but also in how questions are framed, which answers appear credible, and what language is treated as legitimate. The feedback loop: AI-assisted work becomes more visible, more cited, more rewarded, which accelerates the convergence of the very thought that gets fed back into the next generation of training data.

Yes, it’s about architecture, but moreover, homogenization in this case is not a bug in a few writing tools. It is structurally built into the system.

Jay Barney (author of the most-cited paper in strategic management) published a blunt conclusion in MIT Sloan Management Review in 2025: AI will not provide sustainable competitive advantage. His logic is clean. For a resource to generate a durable competitive advantage, it must be Valuable, Rare, and Inimitable. AI is valuable. It is not rare. It is not inimitable. When every function in every organization uses the same models and templates, convergence is not a risk. It is the default.

So is everyone consulting the same oracle, then? Is this just a case of Delphi being overrun and the priests there getting fat on goats and ducks?

Most of the conversation about AI homogenization has stayed in one room: content, brand voice, and communications. Does everything start to sound the same? Yes, measurably. But stay in that one room, and you miss the deeper problem.

Strategy decks drafted with the same prompts on the same three models. HR policies written by systems primarily trained on other companies’ labor laws and risk tolerances. Product decisions filtered through “best-performing” templates harvested from industry averages. Financial forecasts shaped by models that have learned what forecasts are supposed to look like. Sales approaches optimized toward the same conversion patterns everyone else is optimizing toward.

Individually, none of these is catastrophic. Collectively, they push the entire organization toward a single, statistically optimized version of “good,” safe, legible, average. The Deloitte 2026 Global Human Capital Trends report is direct: competitive advantage now depends less on technology differentiation and more on “cultivating the human edge,” adaptivity, creativity, and judgment under uncertainty. Technology is replicable. People aren’t. And the dominant workflow pattern of 2026 is to systematically reduce the influence of what is not replicable.

What we are building, function by function, is an organization that thinks like the average of the internet.

So maybe the question the ancient traditions (and my long-winded way of getting you here) was actually worth asking.

Here is the thing that most frameworks miss: ancient cultures didn’t locate ideas outside humans because they lacked confidence. They did it because it clarified something essential about responsibility.

Plato’s Ion makes this argument inside out. Socrates tells Ion that poets are “interpreters of the gods,” composing not through conscious mastery but through divine possession. This sounds like a compliment. It isn’t. Plato’s point is that if the poem came from the god, not from the poet’s own understanding, then the poet cannot be responsible for its meaning. They are a passive conduit. They cannot explain what they made. They cannot be held accountable for where it leads.

We should all be as suspicious as Socrates was of inspiration precisely because it dissolved authorship.

This is the exact condition we are creating at scale, but with no gods to absorb the blame. The model generates. The employee pastes. The leader approves. When a strategy fails to account for local conditions, when a policy misses its own organization’s culture, when a product answers the wrong question, there is no accountability trail. The AI suggested it. It was statistically probable. It sounded fine.

Nassim Taleb calls this the core ethical failure of modern institutions: separating decisions from consequences. Having skin in the game is not just an incentive mechanism. It is the epistemological condition for understanding the world at all. “If you give an opinion, and someone follows it, you are morally obligated to be exposed to its consequences.”

A model has no skin in the game. It does not attend the town hall after a restructuring. It does not look the client in the eye when a strategy fails. It is not there the morning after.

You are.

The Sumerians didn’t fear the Me. They respected the obligation that came with receiving them. The Egyptian scribe didn’t resent Thoth. They honored the source by doing their craft carefully. The Aboriginal custodian doesn’t own the Dreaming. They are accountable for keeping it intact for what comes next.

In each case, the source of knowledge is larger than the individual. The individual is the point of accountability. Those two things do not cancel each other out. They define each other.

McKinsey’s 2025 global AI survey found that AI high performers are defined not by tool selection but by three things: bold transformation ambitions, intentional workflow redesign, and defined processes for human validation of model outputs. The differentiator is organizational coordination and judgment, not computation.

Barney’s practical conclusion: the advantage is not whether you use AI but what you use it on, your proprietary knowledge, your distinctive context, the judgment at the edge of the situation where the model runs out.

So the question for leaders is not “how do we implement AI across functions?” It is:

In which parts of your organization does human friction — the disagreement, the local variation, the unexpected angle — still have permission to exist?

Because that is where your ideas actually come from. That is where distinctiveness lives. That is where what cannot be replicated gets made. The Aboriginal custodian doesn’t claim authorship of the Dreaming, but is completely accountable for transmitting it faithfully

The ancient traditions were right about one thing across every culture and every century: ideas don’t belong entirely to you. They arrive through conditions, context, and connection, through the specific weirdness of your situation, your people, your moment.

The mistake is not using AI. The mistake is using AI to sand away the very friction that makes your organization worth paying attention to.