A “DELICATE DEAL” that will introduce the first attempt to regulate artificial intelligence in the world was adopted by the European Parliament’s Civil Liberties and Internal Market committee this morning.
The AI act passed by “a very large majority” in its first key vote with the committee. The plenary adoption of the act is expected to be decided on next month before it is negotiated within the EU council and commission.
This is the first law of its kind and attempts to introduce more data, copyright and human rights protections. It is expected to be EU law towards the end of the year.
First drafted in 2021, the act was debated within the committee for exactly a year and amended over, a record-breaking 3,000 times.
The new act will introduce new safeguards to tackle copyright infringement while using AI generated deepfakes through making the developers of AI meet new transparency obligations.
Co-rapporteur of the act, Romanian MEP Dragos Tudorache, said that permission from original authors of copyrighted material will not be required and that developers will train the algorithms just as they train them now.
The new obligations will be able to determine to companies if their content was used in the training phase of a new AI algorithm and Tudorache says developers will now have “an obligation to be transparent about” the material used to train and will be required to “thoroughly document” it.
This will notify the owners of the material and allow them to seek compensation if they see that the products it produces are copying or replicating the original material, he said.
Speaking to reporters yesterday, Benifei said that the act “wasn’t an easy task” and had united people across multiple political spheres.
Tudorache said he thinks that the EU have been working from the very start of negotiations to bring balance in the law between protecting innovation interests and rights of citizens
“At the same time, we also promote this technology. We create the mechanisms within this legislation that would not necessarily put barriers in front of our businesses, big or small,” the Romanian MEP added.
Both MEPs said that this law will increase the level of transparency that EU citizens receive when using AI systems.
Article 52 of the law compels companies to inform the user that they are speaking with a chatbot or if an image or other media has been artificially generated.
According to the MEPs and the law, this will bring a “human-centric” approach to AI, where it is used as a tool to “serve the needs of the people and the common good”.
When asked about how to tackle fake news, MEP Beneifei said “We have the obligations for providers of generative AI that we think will be able to deal with [fake news].”
The Italian MEP said that the specifics of fake news must be “refined” but hopes that Article 52, which makes it a lawful obligation for developers to inform users that the content is AI generated, will be enough to deter malpractices.
Beneifei added that the new act will also make it an obligation for developers to ensure that their AI systems are developed so as to avoid any generation of illegal content.
As well as respecting the choices and freedoms of business, the act will also prohibit practices in AI that are harmful or abusive to EU citizen’s freedom, equality, democracy and rights.
The act will implement the risk-factors, set up by the EU in 2021, ranging from unacceptable to minimal, and will allow developers and engineers to gauge how to introduce their AI product into the European market.
All AI systems categorised in the unacceptable factor, such as harmful or abusive practices will include social credit scoring to anything that could encourage dangerous behaviour in children, will be banned in the EU under the new act.
Tudorache outlined that developers must classify their AI into these factors themselves, and submit their application according to what is required.
The MEP said that companies must assess for themselves if they believe their AI system will “actually pose a significant risk” – which he hopes will stop companies from getting mis-labelled due to keywords within their software.
Tudorache said yesterday that facial recognition technology, that is powered by AI and used on EU citizens will be totally banned, falling under the “unacceptable risk” category.
The Romanian MEP said there was “no stronger safeguard” than to ban the use of FRT completely, particularly in regards to border crossings.
“A border crossing point is a public space. According to the text we have right now, you will not be able to deploy AI biometric recognition technology in a public space,” he said.
The AI Act will also introduce the same safeguards around data protection within the EU and make AI companies subject to the same laws by “ensuring a high level of data protection, digital rights and ethical standards”.
Towards the end of March, Italy temporarily blocked the use of ChatGPT over data privacy concerns, making it the first western country to do so. The service returned to the nation on 23 April 2023.
To embed this post, copy the code below on your site
have your say