An Awkward Audit of the Silicon Orchard

There is a particular brand of quiet, dignified panic that can only be generated within a building constructed entirely of curved glass and extremely expensive concrete. In Cupertino, where the lawns are manicured with what one suspects are tiny, silver-plated scissors, the atmosphere has recently turned rather frosty. Apple, a company that treats the design of its cardboard packaging with the sort of solemn reverence usually reserved for the preservation of medieval manuscripts, has filed a lawsuit against OpenAI. The accusation is one of trade secret theft, conducted, Apple alleges, "at every level." It is a delightful development, akin to a highly refined librarian suing a pirate for borrowing a book without a library card, only to discover the pirate has already melted down the library shelves to build a raft.

To understand the depth of the tragedy, one must first appreciate the nature of Apple’s relationship with secrets. For decades, the company has operated on the principle that a product does not truly exist until it is revealed by a man in a very simple black long-sleeved shirt. Employees are rumored to be divided into small, isolated cells, like members of a particularly polite resistance movement, none of whom are permitted to know what the others are doing. It is entirely possible that the person designing the volume button on the next telephone believes they are working on a highly classified guidance system for a submarine. To have this carefully curated silence breached by OpenAI—a firm whose foundational philosophy appears to be that if a piece of data is not nailed down, and indeed even if it is, it is technically public domain—must feel like a personal affront to the very concept of property.

The lawsuit claims that OpenAI systematically harvested Apple’s proprietary technology to construct its own upcoming suite of artificial intelligence gadgets. One can only imagine the scene of the alleged crime: a rogue engineer, perhaps wearing a slightly less minimalist turtleneck, transferring folders with names like "Extremely Secret Rounded Corners" or "Slightly More Sincere Siri Voice" onto a USB stick disguised as a piece of artisanal cheese. The legal documents describe a scheme of breathtaking scale, suggesting that OpenAI’s engineers did not merely copy the homework, but actually moved into Apple’s spare bedroom, wore Apple’s slippers, and began answering Apple’s landline.

There is, of course, a magnificent irony at play here. OpenAI has spent the last several years explaining to various courts, authors, and outraged painters that the act of taking things from the internet and running them through a very large blender is not "theft," but rather "learning." It is a defense that relies on a charmingly optimistic view of cognitive development. If a human child walks through an art gallery, looks at a hundred paintings, and then paints a picture of a horse, we do not sue the child. We praise them, even if the horse has five legs and looks slightly like a bicycle. OpenAI argues that its algorithms are merely very large, very expensive children. Apple, however, is not inclined to be a proud parent. They view the situation more as a case of someone entering their orchard, eating all the apples, and then selling a slightly bruised pear back to them at a premium.

I once knew a gentleman in Gloucestershire who spent three years constructing a remarkably complex grandfather clock out of discarded biscuit tins. He was entirely convinced that the local postman was spying on his progress through the keyhole with the intention of mass-producing biscuit-tin clocks in Switzerland. When the clock was finally finished, it did not tell the time, but instead emitted a low, rhythmic hum that attracted wasps. The postman, when questioned, admitted he had only ever been looking for a pencil. One wonders if the "trade secrets" Apple is so fiercely defending are of a similar nature—highly complex, beautifully engineered, and ultimately destined to tell us that the weather in London is currently grey, which we could have ascertained by looking out of the window.

The legal battle promises to be a spectacular clash of corporate cultures. On one side, we have the disciples of the Walled Garden, who believe that the universe should be clean, seamless, and require a proprietary charging cable. On the other, we have the champions of the Infinite Scrapbook, who believe that the universe is a giant, unstructured database waiting to be indexed and monetized. It is a conflict between the architects of the cathedral and the people who want to turn the cathedral into a very large, very noisy server farm. The court will be asked to decide where a "secret" ends and a "statistical probability" begins. If an algorithm predicts the next word in a sentence based on a document it was not supposed to read, has it committed espionage, or has it simply been a very attentive listener?

During a brief and largely unsuccessful tenure as an assistant clerk for a small maritime insurance firm, I observed a dispute over a cargo of nutmeg that had allegedly been "influenced" by the proximity of some highly aromatic onions. The insurers argued that the nutmeg was no longer nutmeg, but had become a new, onion-adjacent entity. The merchants insisted it was still nutmeg, merely nutmeg with a broader perspective. The court eventually ruled that the nutmeg should be sold to a soup manufacturer, who did not care either way. One suspects the ultimate fate of these disputed AI models will be much the same—they will be absorbed into the great, grey soup of consumer technology, where they will be used primarily to draft emails apologizing for being late to meetings.

In the meantime, the rest of us must look on with the mild, detached amusement of spectators at a croquet match where the players have begun using their mallets to settle personal grievances. Whether Apple wins or OpenAI successfully argues that they "accidentally" learned Apple's secrets while looking for something else, the result will be the same. We will continue to receive devices that are slightly thinner, slightly warmer, and significantly more opinionated about how we should live our lives. And Siri, presumably, will remain just as confused as before, though perhaps with a slightly more sophisticated vocabulary stolen from a very expensive legal brief.