AI and Chariots of Fire

In 1924, a Cambridge sprinter named Harold Abrahams committed a scandal: he hired a professional coach named Sam Mussabini. Mussabini helped Abrahams win Olympic gold. If you've seen the film Chariots of Fire, you know how the Cambridge establishment reacted. Coaching, they sniffed, was corrupting. Amateur athletics is supposed to display the natural born technique that is a sign of innate character. A gentleman would never accept hired expertise.

A century later, that objection sounds both naïve and offensive. Still, it resembles a debate we're having today about the use of AI. I wrote a commentary about it for the Journal of Medical Ethics, responding to a proposal by Brian Earp and colleagues that scholars should be required to disclose whether and how they used generative AI in their manuscripts. The parallel is not exact but suggests the ways we think about human capabilities. Questions about the use of AI as a writing coach are, structurally, similar to the anxiety that Cambridge dons had about Abrahams' running coach.

Both debates turn on questions about what we mean by human excellence, and what counts as cheating in the pursuit of it.

The purity worry, and why coaching complicates it

There is a real concern about writing and AI. Writing isn't just a way to produce text. At its best, writing allows us to refine and clarify our own ideas. We can discover our arguments by struggling to make them and by finding their shortcomings. If AI gins up a polished manuscript and you skip that struggle, you may end up with something technically superior but not really yours.

Good coaching does something different. Mussabini didn't let Abrahams sit and watch while his coach ran the race. Instead, he helped Abrahams understand how to improve his stride. As a result, Abrahams was able to run faster, faster than anyone else in the world that year. The distinction between coaching and outsourcing matters. It is the difference between work that develops one's capacities and assistance that substitutes for them. A model that flags a weak argument, proposes an objection, or shows where logic fails is coaching. A model that writes the paper while you nap is running your race.

Here's the problem: most disclosure statements can't help a reader tell those two apart. "AI was used substantively" could mean it caught a typo or it wrote your first draft. Saying only that AI was used tells a reader about as much as saying an athlete was coached.

Where disclosure could backfire

Chariots of Fire suggested that the concerns about coaching were not merely about athletics. Abrahams wasn't just a man with a coach. He was a Jew, the son of immigrants, an outsider in a racist, stodgy Cambridge. His coach was of Syrian and Italian descent. The Cambridge masters' disdain for the coaching was tangled up with their disdain for the men themselves. They dressed up bigotry as a defense of amateur purity. Then, as with many debates about insiders and outsiders, they accepted Abrahams as one of their own once he won gold. Qualms about coaching quietly evaporated.

Disclosure of AI assistance could also cut two ways. AI can be an equalizer: a brilliant scientist with weak English can finally compete on the strength of ideas rather than idiom. But if we flag AI use and non-native speakers lean on these tools most to reach field-standard prose, then a policy meant to normalize transparency could become discriminatory. We'd be reproducing the old suspicion of the outsider who needs help to compete.

What is an author?

Foucault argued long ago that the "author" was never simply the person who produced the words. Authorship is a signal that shapes how we read a text. The fight over AI declarations is a small piece of a much larger renegotiation of what it means to be smart, articulate, accountable, creative. Or necessary.

My bet is that today's concerns will soon seem as quaint as Cambridge's horror at a running coach. The real measure of a good disclosure policy may be its eventual disappearance. Responsible AI assistance is already ordinary. Irresponsible work should be rejected for its defects, whatever produced it. In the end, papers should be judged by the quality of their arguments and evidence, not by the type of human-machine collaboration that made them.

Abrahams broke the tape. He was still the one who ran fastest.

Lantos JD. Generative AI and Chariots of Fire. J Med Ethics. Published online September 15, 2026. doi:10.1136/jme-2026-112333.