Europe inches closer to first-ever AI Act: What it means
What's the story
The AI boom has been exciting for many reasons. However, the increasing popularity of AI-powered chatbots has raised questions about their social, economic, and security impacts.
People have been calling for AI regulations. Now, Europe is inching toward its landmark Artificial Intelligence (AI) Act.
A group of European lawmakers has reached a preliminary agreement on the act. What does it mean for AI companies?
Context
Why does this story matter?
The list of AI-powered chatbots is getting longer day by day. We have even seen glimpses of AGI with Auto-GPT.
These chatbots are capable of mimicking human beings, excelling at exams, ordering food on your behalf, and a lot more. However, they are not without flaws.
Lawmakers believe AI must be regulated before it's too late.
Regulation
EU first proposed the regulation 2 years ago
The European Commission first proposed the AI Act two years ago. It was a move meant to protect people from the harms of the emerging technology.
ChatGPT's entry, however, popularized AI. What followed was a flurry of investments and more chatbots, prompting the EU to hasten the push for the legislation.
Now, the world's first comprehensive legislation governing AI is close to being passed.
Copyright
Companies will have to reveal the usage of copyrighted material
What does this mean for AI companies and generative AI? One of the things companies would be apprehensive about is the requirement to reveal copyrighted material.
AI models like GPT-4 and LaMDA are trained on troves of data, which ought to include some copyrighted material.
If the bill is passed, companies will have to disclose whether they used copyrighted material in their AI development.
Classification
The act classifies AI systems into 4 tiers
The proposed legislation is an all-encompassing one. It classifies AI based on the risk it poses to the health and safety or fundamental rights of a person.
There are four tiers: minimal, limited, high, and unacceptable. The act covers anything from spam filters or video games to government social scoring and general purpose AI like ChatGPT.
Depending on the risk, regulations change.
Compromise
European Parliament calls the act a 'solid compromise'
There are many who see regulating AI as a hindrance to the technology's development, while there are others who think the proposed legislation isn't enough.
Svenja Hahn, a European Parliament deputy, said the AI Act is a "solid compromise that would regulate AI proportionately." She said it protects citizens' rights without hampering innovation.
It needs to be seen how this compromise works out.
Next step
The draft has reached the 'trilogue' stage
A group of MEPs (members of the European Parliament) voted in favor of the AI Act on Thursday. The draft has now reached the next stage called 'trilogue.'
These informal meetings between MEPs, the Council, and the Commission help reach a provisional agreement on acceptable text to both the Council and the Parliament.
Then, there will be formal procedures to approve the text.