Generative AI wasn’t featured on the EU lawmakers’ plans for restraining AI technologies such as ChatGPT. However, by April, associates of the European Parliament (MEPs) were looking forward to those rules to catch up with the interest in generative AI, which has provoked anxiety after the release of AI chatbot ChatGPT.
- The word “chatbot” in Bloc’s 108-page proposal AI act is referred to as Deepfakes: Images and audio that mimic humans.
- A new draft of legislation was issued to help identify copyright protection prominently to keep Artificial Intelligence in check.
Generative AI was not featured prominently in the EU lawmakers’ plans for restraining AI technologies such as OpenAI’s advanced chatbot ChatGPT. Bloc’s 108-page proposal for the AI Act was issued 2 years ago, and the proposal mentioned the word “chatbot” only once. The references towards AI-generated content were referred to as Deepfakes: audio or images designed to mimic humans.
However, by April, associates of the European Parliament (MEPs) were looking forward to those rules to catch up with the interest in generative AI, which has provoked anxiety and awe ever since ChatGPT was first released in November.
On Thursday, a new draft of the legislation was issued to help analyze copyright protection prominently to keep Artificial Intelligence in check.
For the first time, the interviews revealed the four lawmakers and two close sources to discussions.
A group of politicians was able to hammer out what might become landmark legislation over just eleven days and reshape the regulatory landscape for OpenAI along with its competitors.
For now, the draft has not been concluded and according to lawyers, it might take a few years for it to be in force.
LAST MINUTE CHANGES
Ever since ChatGPT was launched back in November 2022, it became one of the best AI chatbots in the AI market. This also results in other major tech giants creating an AI chatbot of their own and investing in generative AI startups such as Midjourney and Anthropic.
The massive popularity and usage of AI chatbots like OpenAI’s ChatGPT have resulted in Thierry Breton, EU industry Chief, and others requesting for restrictions to be placed on these AI chatbots.
The organization backed by Billionaire and CEO of Tesla and Twitter, Elon Musk, took it the notch by stating a letter warning of existential risk from Artificial Intelligence and requesting stricter regulations.
Various MEPs involved in the drafting and legislation signed the open letter and agreed to some parts of Musk’s open letter on 17 April. Urging the world leaders into holding a summit to find methods to control the further development of AI.
Although on the same day, two of them, Brando Benifei and Dragos Turdorache, proposed a few changes that would be required for companies with generative AI systems to reveal any copyrighted material that has been utilized while training the models.
The anonymity was requested according to four sources present due to the sensitivity of the discussions. The new proposal received the support of the cross-party.
A proposal by Axel Voss, a conservative MEP, was requiring companies to ask for approval before using the data or information from rightsholders. Although this proposal was rejected due to being too restrictive and raising chances to stifle the growing industry.
The laws or rules presented by the EU tabled were made to reach an uncomfortable level of transparency in the secretive industry.
Tudorache stated, “I was quite surprised at how efficiently we converged on what should be in the text on these models”. Showcasing consensus and shared understanding.
The voting by the committee on the deal will take place on 11 May and if it’s successful the process will be moved forward to the next stage of negotiations, the trialogue. In this stage, the members of EU states will be discussing the content with the European Commission and parliament.
BIG BROTHER AGAINST. THE TERMINATOR
MPs were unconvinced until recently that generative AI deserved any sort of special attention.
In February, Tudorache stated, generative AI was not going to be covered in depth.
Quoting concerns over data protection over chatbot warning, he stated, “I’m more terrified of Big Brother than Terminator”
However, Tudorache did state he and his colleagues know that generative AI requires laws that specifically target its usage. Among new proposals that target “base models”, software companies such as OpenAI backed by Microsoft, would have revealed any copyrighted material such as articles, books, blogs, images, videos, and more to train their AI models.
In recent months, allegations of copyright infringement have rocked among the AI companies with Getty Image suing Stable Diffusion to OpenAI being criticized for refusing to share details of their datasets.
MEP Svenja Hahn stated, “Various calls have taken place from outside and inside Parliament regarding placing a ban or classification of ChatGPT at high risk.”
The final compromise is favorable to innovation since it doesn’t classify these models to be at ‘high risk’ but instead sets requirements for quality and transparency.