EU's New AI Rules: A Peek into the Future

The European Union is taking a big step towards regulating AI with its draft rules for general-purpose AI models. Big names like Meta, Google, and OpenAI will need to follow these guidelines. You might be wondering, what are these general-purpose AI (GPAI) models anyway? They're like super-flexible AI that can handle many tasks without much adjustment. The EU's draft is like a roadmap for these companies to stay on the right track and avoid hefty fines. The draft is all about playing it safe. Companies need to be open about how they train their AI, especially those web crawlers that gather data. It's also about making sure AI doesn't go rogue or cause harm. There's a special focus on cybersecurity and preventing discrimination. Companies will have to show how they're managing these risks and have a backup plan in case things go wrong. The EU wants these AI developers to be accountable and keep checking for risks. They might even need to bring in experts from outside the company. The penalties for not following the rules are pretty steep – up to €35 million or 7% of a company's global annual profits. Stakeholders have a chance to share their thoughts on the draft until November 28. The final rules are expected to be ready by May 1, 2025. Some big tech companies aren't happy about these rules, but others support them. It's clear that AI development needs to be watched closely.