A bill to oblige disclosure of AI learning data is submitted in the EU
A bill to oblige companies to disclose AI learning data has been submitted in the EU, and it has become clear that it has taken a step towards enactment.
EU proposes new copyright rules for generative AI | Reuters
Europe to ChatGPT: Disclose Your Sources - WSJ
https://www.wsj.com/amp/articles/europe-to-chatgpt-disclose-your-sources-863ef330
According to the new bill, companies that develop text-generating AI ChatGPT, image-generating AI Stable Diffusion, Midjourney, etc. will be obliged to compile the learning data used in the development of the system and disclose the details of the data. That's what I'm talking about.
AI can produce highly accurate content by learning billions of sentences, images, and videos that are flooding the world, but the data used for learning includes copyrighted content. It is often pointed out that there is a possibility that
In light of this situation, the EU has included a clause in the AI bill drafted in 2021 that requires the disclosure of learning data. This bill will be reviewed again in April 2023, and it has been decided to proceed to the next stage toward enactment.
If the bill is passed, it will be a boon for creators who want to know how much their content has been learned. If it turns out that copyrighted content has been studied, the author may be able to seek profit sharing or even take legal action.
In the EU, where corporate data collection is strictly regulated under the General Data Protection Regulation (GDPR), Italy has
Under GDPR, tech companies must obtain user consent before using personal data to train their products. Companies doing business in the EU should also give EU citizens the choice to opt out of data collection and sharing.
The Italian data protection authority that banned ChatGPT requires the developer OpenAI to comply with Italian and EU laws by April 30, 2023. However, AI experts say, ``Since OpenAI's models are trained on a large amount of information gleaned from the internet, even developers cannot know exactly what data is being collected. would,” he pointed out. It is said that it is almost impossible for OpenAI to meet the demand.
ChatGPT's 'OpenAI', which may have been trained without user consent, faces legal problems in the EU that emphasizes strict privacy laws, experts say 'compliance with rules is nearly impossible'-GIGAZINE
The new bill will regulate organizations that manage all generative AI, including ChatGPT, and require them to be highly transparent about their AI operations.
According to Reuters, some committee members initially proposed a complete ban on the use of copyrighted content for learning, but this was abandoned in favor of transparency requirements. It was said that it was done. Vice-President of the European Parliament Svenja Khan said: 'Parliament has found a firm compromise to regulate AI proportionately and protect the rights of citizens, while promoting innovation and stimulating the economy. I said.
Related Posts:
in Software, Posted by log1p_kr