The EU AI Act Enacted: Mistral Celebrates | The AI Beat

The EU AI Act Enacted: Mistral Celebrates | The AI Beat

Stay updated with our daily and weekly newsletters to get the latest insights and exclusive content on leading AI developments.

After three days of intense negotiations, the EU AI Act was finally approved over the weekend. Carme Artigas, Spain’s secretary of state for digitalization and artificial intelligence, hailed it as a “historical achievement and a huge milestone towards the future!”

The EU AI Act mandates that ‘high-impact general purpose AI models’ must comply with transparency standards, while ‘high-risk’ models face additional requirements such as risk management, severe incident monitoring, AI model evaluation, and red teaming strategies.

Initially, the AI community celebrated the Act, but excitement waned once it became clear that it’s only a provisional agreement. Final wording could take months, and enforcement won’t begin until two years after the European lawmakers give their final approval. In the fast-evolving world of AI, two years might as well be two centuries.

The spotlight on the EU AI Act quickly dimmed when Mistral, a Paris-based open-source model startup, made headlines. Founded just seven months ago by former Meta and Google researchers, Mistral unveiled a new LLM using only a torrent link. The following day, they announced raising $415 million from investors like Andreessen Horowitz, Lightspeed Venture Partners, Nvidia, and Salesforce, valuing the 22-person company at about $2 billion.

Mistral has played a significant role in the EU AI Act negotiations, advocating against the European Parliament’s tiered regulation for generative AI. Cedric O, France’s former state secretary for digital affairs, led Mistral’s lobbying efforts, stressing that the AI Act could either cripple or aid Mistral, depending on its final form. Eventually, the Act included broad exemptions for open-source models, except those posing a systemic risk, marking a lobbying success for Mistral.

TechCrunch aptly noted that Mistral’s success highlights the struggle for AI sovereignty in Europe. The EU aims to lead technologically while regulating the industry to promote safe development. Germany’s Vice Chancellor Robert Habeck emphasized that having strong AI regulations but no European companies would be a hollow victory for the continent.

AI regulation will likely be attempting to catch up for a long time, whether in the EU, the US, or elsewhere. As Mistral’s Mixtral 8x7B model outperformed OpenAI’s GPT 3.5, thousands of AI researchers gathered at NeurIPS in New Orleans to discuss generative AI advancements. There, they’ll toast to rapid AI progress in 2024 while policymakers worldwide strive to balance innovation and regulation. If Mistral’s recent triumph is any indication, regulators urgently need to pick up the pace.