Op-Ed: AI, plagiarism, law, sanity, money, and not doing much about any of it

Chinese schoolchildren are using ChatGPT to slash homework time, but teachers are worried over the possibilities for cheating and plagiarism – Copyright AFP Ed JONES

The fully justified fuss about AI plagiarism is an allergic reaction to a very real set of problems. What’s not happening is any sort of systemic shut down of the problems.

Plagiarism comes in two basic forms. There’s actual copying and tweaking of copyright content. They’re effectively the same. The original content is stolen and then on sold as original by the people who stole it There’s nothing easier for AI, which uses machine learning and neural networks to process images, text, and any sort of media.

Legally there are no outs for plagiarists. This IS theft.

That’s part of the problem; the people doing it. Plagiarism is not exactly new. Thieves are the world’s least useful people. Plagiarism, like crime, is a perfect career for people who generate nothing of value themselves.

To make it slightly more infuriating, they think they’re being “clever” when they steal. That’s a big incentive, an ego-builder for people who are otherwise nobodies. Take that away from them, and plagiarism will be much less of a problem.

It’s an expensive and destructive problem for content creators, publishers, and consumers. Consumers suffer from the fact that their sources don’t actually exist. The plagiarized content could be garbage. AI is predicted to replace fake news any day now.

(A moment’s silence for all those soon-to-be unemployed trolls and political skanks if you please. Then break out the champagne.)

For content creators, the problem of plagiarism is a bottom-line thing. It eats directly into their ability to earn income. Their rights are automatically infringed.

For publishers, there’s a very clear and eventually inevitable expensive legal issue with publishing another publisher’s content. It’s stealing their business, as well as their property.

Despite some noticeable legal actions regarding sources of copyright materials for AI learning, there’s no standout benchmark case for plagiarism. There should be a straightforward “this is what happens if you plagiarize” case. What’s needed is a famous legal kick in the teeth as a turnoff for the AI-obsessed idiots. This is big business and potentially huge money for claimants against plagiarists. That definitely will get attention from the large end of town.

There’s another inherent problem. AI is still an unknown. People are scared of AI. They don’t really get it. If they did, they’d be a lot less scared of it and much more aggressive when responding to plagiarism.

For content creators, learn more about AI. It’s actually pretty straightforward. You will see ways of securing your work. This YouTube video may be a bit heavy for non-maths people, but it explains the basics of how it works, how it learns and what you can do about it.

You’ll note that “recognition” includes elements you can customize to personalize your work. In theory, you can even build in watermarks, code and micro routines to scream “Plagiarized content! onscreen” that should scare oft the parasites fairly effectively. The joke is that most of them are largely code-illiterate themselves.  

There are many ways of managing AI content that already exist. One of them is shutting them down at the source. A recent case of “parasitic AI”, publishing tweaked content included anonymous registration of domains. You can see that problem easily enough.

ISPs can be and often are seen as enablers for legally actionable matters. It’s a version of “vicarious liability” that may or may not stand up in a court. That’s not a legal position anyone wants. It’s also not good enough as a definition of who’s liable for what. That needs to change.

Anonymous registration? Why? You don’t have to register anything if you don’t even know who you’re doing business with. As an ISP, you could be taking on a massive liability. Why would you do that?

Terms of service for AI domains could include anything from no AI-generated content to acceptance of liability for any legal issues, etc. Even the Dumbest of the Dumb know a major risk when they see one, sooner or later.

Slight digression here, but it’s relevant.

In the case of a site that ran plagiarized content, the AI tweaked the original content as follows:

“We believe he’s only just scratched the surface of his potential.”

It was re-written as the nonsensical:

“He’s nudged the iceberg but there’s so much more to come.”

The “iceberg” reference was botched, but it’s an idiomatic tweak, not just substituting equivalent terms. It retains the sense of the original using hardly any of the original text. That has some ramifications for managing plagiarism.

It’s a bit more interesting and complex than it looks, and far more problematic. If AI can tweak it like that, the plagiarism is a bit better covered. It’s still full of holes, notably structurally, but it’s less obvious.

This means that in the future, legal rules for plagiarism will have to include the idioms and contexts of content as assessments of the plagiarism. This is a sort of contextual search, but you know how good those are. You may have to go from Search Engine Optimization to Search Engine Overkill.

It’d be a lot easier to create a simple system for content creators where original content is verified. Any use other than the standards of “fair use” for media could be considered plagiarism or wannabe plagiarism.

Verification systems already exist in publishing. The content creator ticks a few boxes and is allowed into the system as an authentic user. Pretty much ironclad. You could use an SSL system if you want to be meticulous.  

This also adds a level of difficulty for plagiarists if anyone takes the trouble to do a CAPTCHA-like system for scanning content. AI can do that, too. As a way of spotting plagiarism, it’s the perfect tool.

These thieves have to go. Whatever it takes needs to be done.