Negotiations are reportedly underway within the European Union (EU) to formulate additional regulatory measures specifically targeting the largest artificial intelligence (AI) systems. The discussions involve representatives from the European Commission, the European Parliament, and various EU member states. The focus is on addressing the potential impact of large language models (LLMs), such as Meta’s Llama 2 and OpenAI’s ChatGPT-4, and determining whether additional restrictions should be imposed on them. These discussions are part of the broader context of the forthcoming Artificial Intelligence Bill.
According to insiders cited by Bloomberg, the primary objective is to strike a balance between regulating large enterprises that employ LLMs and avoiding excessive regulatory burdens on emerging startups in the AI sector. Negotiators have yet to finalize an agreement on this topic, and the discussions remain in the preliminary stages. It's worth noting that the EU's approach to AI regulation is expected to mirror its Digital Services Act (DSA), which aims to establish standards for user data protection and content monitoring on online platforms and websites. However, more stringent controls are likely to be imposed on the largest online platforms.
Companies like Alphabet and Meta are among those that will need to update their practices to comply with the EU's new standards under the DSA, with a deadline set for August 28. The EU's Artificial Intelligence Bill represents one of the earliest sets of mandatory rules for AI developed by Western governments. In contrast, China has already enacted its own set of AI regulations set to take effect in August 2023. Under the proposed EU regulations, companies involved in developing and deploying AI systems will be required to conduct risk assessments, label AI-generated content, and, notably, ban the use of biometric surveillance, among other measures. However, it's important to note that this legislation has yet to be enacted, and member states still retain the ability to disagree with any proposals put forth by the parliament. In China, over 70 new AI models have reportedly been introduced since the implementation of its AI law.
