Alessandro Nodari of 26 May 2023, 1:07 pm
Artificial intelligence is here to stay, but that doesn’t mean theEurope Union has no intention of regulating the use of ChatGPT and partners. To the point that Sam Altman himself, head of Open AI, claimed to have “many worries” on the law and to be able to consider the withdraw of its services from the EU.
The threat is the same one aired by Mark Zuckerberg in 2022, when pressed by the EU he threatened to abandon the Continent, and then backtrack. One thing is certain: no one can do without the European market, whether Zuckerberg or AI.
But what’s going on? The EU is reportedly finalizing the AI Act to require companies working on AIs to disclose i details their training methods and data sources. This in the hope of making more transparent not only the development, but also any copyright attributions, a problem that the advent of Midjourney has recently raised.
The terms of the law have been expanded in recent months to include new obligations for creators of so-called “base models,” large-scale AI systems that power services like ChatGPT and OpenAI’s DALL-E.
But Altman said how these rules are difficult to be respected, and which might even consider the withdraw of its services from the European market in response to AI regulation. The entrepreneur had told Time that his concern was that systems such as ChatGPT would be designated by the EU as high risk.
In addition to the technical challenges, the disclosures required by the EU AI Act also have potential commercial threats for OpenAI. A provision in the current draft requires creators of base templates to disclose details about design of their system (including “required computing power, training time, and other pertinent information regarding model size and power”) and provide a “summary of copyrighted data used for training“.
OpenAI initially shared this type of information but then stopped as its tools became more and more precious from a commercial point of view.
Not to mention the fact that forcing OpenAI to identify its use of copyrighted data would expose the company to potential lawsuits. Using data from the Web, many actors could ask for rights, as it did Getty Images for using his copyrighted data to train the AI image generator. A bit like saying “everyone knows that we use other people’s data, but it’s better not to write it in black and white to give it a legal foothold to ask us for money“.
Evidently, however, Altman has changed his mind, because after a series of conversations with the European legislators it has opened a window. On the contrary, “he has no intention of leaving“.
What will they have said? The details have not been leaked, but if on the one hand Altman certainly cannot abandon a market like the European one, it is possible that they are better off outlined the features of the legislation, in order to propose it as less limiting for OpenAI.
Or maybe he was just talking to his investors (anyone mentioned Microsoft?).
At present, the company is also the focus of a number of comparisons with US law, but while in the US regulation should apply primarily to future systems, the EU AI Act takes care of present and current AI capabilities.