Feeding the Beast: Are you still giving your ideas away for free?

It’s a provocative question but one that you’ll appreciate being asked.

In the race to publish, share, and be seen, researchers are now facing an uncomfortable new reality: the more accessible your work is, the more likely it is to be consumed, not by peers or students, but by a machine.

Large Language Models have created an appetite for text that knows no boundaries. Trained on vast volumes of content scraped from the open internet, including academic articles, blogs, preprints, and reddit comment sections, they operate with a simple directive: absorb everything, attribute nothing.

It’s the silent digitisation of intellectual labour. Packaged, paraphrased, and served back to the world with the shiny packaging of fluency but none of the depth, nuance, or credit.

The Illusion of Protection

Many researchers have found comfort in open access models, trusting that Creative Commons licences or institutional repositories offer a layer of protection. In practice, these protections are illusory. Current LLMs don’t honour licensing agreements. They don’t differentiate between a peer-reviewed publication and a half-baked forum post. They simply train on what they can access.

This is not an attack on open science. The goal of knowledge sharing remains noble. But we must now interrogate the infrastructure through which that sharing takes place. Because when your work is fed into an AI system, it doesn’t remain yours. It becomes only data. Disembodied, decontextualised, and indistinguishable from everything else.

Publishing vs. Control

There’s a critical distinction between publishing your ideas and preserving control over them. Publishing invites peer review, discussion, growth. It operates in a system of norms—citations, acknowledgements, and professional ethics.

Feeding your work into the digital commons, where LLMs forage, is a different proposition. These systems strip content of provenance. An insight hard-won through years of analysis might resurface as a bullet point in a chatbot's reply, unattributed and distorted.

If that feels uncomfortable, it should.

The Machine Forgets Who You Are

The core issue here isn’t that AI tools use your ideas but that they forget who you are.

Attribution in academia isn’t etiquette. It is the mechanism by which careers are built, credibility is established, and intellectual lineage is traced. When machines bypass this entirely, they erode the identity of the researcher, AND the whole dang structure of scholarly discourse.

You did the fieldwork. You flirted with insanity over the transcripts. You coded the data, challenged your assumptions, and crafted an argument. That work deserves better than to be flattened into a statistically probable sentence.

Reclaiming Your Intellectual Sovereignty

In this new landscape, knowledge must be treated not as fodder but as property. Something earned, held, and consciously distributed.

That doesn’t mean locking everything behind a paywall. It means asking better questions before we hit ‘publish.’ It means being intentional about where our ideas go and who gets to use them.

Before You Upload That Next Preprint, Ask Yourself:

• Who will benefit from this being online?

• Am I comfortable with this work being ingested by AI systems without credit?

• Are there controlled ways to share this insight, within a trusted network or community?

• If I were cited in every reuse of this idea, would I still feel uneasy?

• What part of this knowledge do I want to remain mine?

We are living through a shift in how knowledge is valued, consumed, and attributed. The systems we build and the choices we make determine whether researchers retain agency in that future, or whether our work is merely fuel for a machine that forgets where it came from.

The ideas are your own. You should get to decide their destiny.