If you are not sure of the subtleties between copying, infringing or being inspired, ChatGPT can explain all.

 

More than that: ChatGPT has become the guardian of intellectual property and personality rights.

Here is an example touching on the personality rights of Kate Moss.

 

I think its fair to say tha Kate Moss has nothing to fear !

When it comes to imaginary characters, this conversation shows the level of sophistication going on:

 

When asked what makes this Pixar style, ChatGPT gives you a list of 5 characteristics:

• Expressive
• Vibrant lush colours
• Detailed environment
• Whimsical tone
• Softt lighting and depth
• Family oriented theme

Going one step further, I ask ChatGPT to create a character which “Looks like” the character Anxiety in the Pixar Movie Inside Out. Here is the result:

 

No visual similarity between this image and the original.

Now, going one step further I ask :

 

ChatGPT clearly puts its foot down: copyright infringement is not on under its guard.

Ask ChatGPT to explain how it delineates the boundaries between infringing copy and free inspiration, it will give you a full course on copyright infringement and even offers a concrete example:

In its infancy, ChatGPT was not so careful about the output it generated. So why is it so explicit

in identifying and respecting the copyright of protected works?

It’s certainly a central element in its defense strategy in the many law suits it faces.

Open AI faces two main attacks based on IP infringement:

• The first claim is that the scraping and tokenisation of protected works to train the AI is equivalent to an infringing reproduction;
• The second is that the output generated by ChatGPT will infringe copyrighted works since it feeds off these works

Of the two claims, the first certainly strikes to the heart of OpenAI. Today, it feeds off protected content for free … and content is key.

Being able to establish that guardrails have been put in place to prevent a user from generating a copyrighted content is a clever way of proving to the courts that the tokenisation process is harmless in itself and does not result in copyright infringement. It is a way to reinforce the anthropomorphic idea that just as a human creator finds inspiration in what already exists, so the AI feeds off the existant to produce something original. More than that, the superior intelligence of the AI will guarantee that the output is not infringing – something a human canno guarantee as he/she may have been inspired from a protected work without realising it.

One of the many cases pending in the US has just been settled. The reference to the “guardrails” put in place to avoid infringement is central to the agreement… and in all fairness, ChatGPT is now certainly educating the wider public to the basics of IP all the while limiting the risk of innocent infringement.

***

But does this address the question of the inherent value of the content OpenAI is feeding on daily to give substance to its outputs? Does it answer the question of why OpenAI should be able to offer a commercial service which relies on the work of content cretors without remunerating them ? No.

Perhaps copyright infringement is not the best legal ground after all ? Perhaps notions of unfair competition stand a better chance before courts. After all, rather than let OpenAI use a copyright claim to its advantage, a real debate on the notion of fairness outside all copyright considerations might put Open AI in a more uncomfortable spot.

Simply put : is it fair that Open AI should “steal” intangible value in which third parties have invested money and effort over time, to commercialise a content-generating service the value and quality of which depends entirely on the content it has trained its programme on?

And so, those sophisticated guardrails feed into the notion of fair use – Open AI’s primary defense.

More than that, these guardrails are the living proof of Open AI’s declared mission “to make AI benefit humanity”.

Socratic syllogistic logic at its best!

Olivia Dhordain
11.01.2025