EU is investing $56m to enter the AI race

The European Commission has announced a €54($56) million investment in OpenEuroLLM, an open-source large language model aimed at competing with U.S. and Chinese AI models like ChatGPT and DeepSeek. Developed by over 20 European companies, universities, and supercomputing centers, the AI will support 35 languages and be accessible to businesses, governments, and citizens.

While the initiative is significant, its funding is minimal compared to leading AI labs like OpenAI ($21.9B raised), Anthropic ($9.7B), and Mistral AI ($1.2B). The EU’s funding is, however, 10 times what DeepSeek claims to have spent training its R1 model, suggesting that a breakthrough is still possible.

The project aims to ensure alignment with EU values and legal frameworks, especially in response to restrictive AI regulations that have hindered U.S. tech companies from operating in Europe. Despite past failed European tech projects and concerns about inefficiency, this initiative could be a step toward AI self-sufficiency for the EU.

Linked to full article
OpenEuroLLM website

OpenAI introduce new AI Agent Deep Research

OpenAI has introduced Deep Research, an advanced AI agent designed to autonomously conduct multi-step research on the internet, significantly reducing the time required for complex tasks. Available now for Pro users, it will soon expand to Plus, Team, and Enterprise subscribers.

Key Features:

• Uses reinforcement learning to plan and execute research tasks.

• Synthesizes large amounts of online information, including user-uploaded files.

• Creates visualizations using Python and embeds relevant graphs and images.

• Cites specific sources, improving transparency and reliability.

Limitations:

• May still hallucinate facts or misinterpret authority of sources.

• Has formatting and citation errors, though improvements are expected.

• Longer response times due to high computational demands.

Link to full article

EU AI Act Takes Effect: Unacceptable AI Systems Face €35 Million Fines

The EU’s AI Act, which officially took effect on August 1, has reached its first compliance deadline on February 2. Under this framework, regulators can ban AI systems deemed to pose an "unacceptable risk," with violations carrying fines of up to €35 million or 7% of annual revenue.

The Act categorizes AI risk into four levels: minimal, limited, high, and unacceptable. Unacceptable applications—now banned—include social scoring, biometric-based profiling, emotion inference at work or school, real-time biometric data collection for law enforcement, and predictive crime analysis based on appearance.

Many companies, including Amazon and Google, have voluntarily pledged compliance through the EU AI Pact, but notable players like Meta, Apple, and Mistral have not signed. Enforcement will ramp up in August when competent authorities are designated.

Certain exceptions exist, such as law enforcement’s use of biometric data for targeted searches or safety-related emotion analysis in workplaces. Additional compliance guidelines are expected in early 2025, though questions remain about how the AI Act will interact with existing regulations like GDPR and NIS2.

Link to full article

Heritable Agriculture, a new Google Startup for AI for Crop production

Heritable Agriculture, a newly independent startup spun out from Google’s X “moonshot factory,” leverages machine learning to optimize crop production. Founded by Brad Zamft, a former Google X project lead, the company analyzes plant genomes to identify optimal breeding combinations that enhance yields, reduce water consumption, and improve carbon storage. Unlike gene editing, Heritable focuses on traditional breeding techniques.

The startup tested its AI-driven models on thousands of plants in growth chambers and field sites across the U.S. Agriculture, a major contributor to greenhouse gas emissions and resource depletion, stands to benefit from Heritable’s data-driven approach.

Now focused on commercialization, Heritable has secured funding from FTW Ventures, Mythos Ventures, SVG Ventures, and Google, which retains an undisclosed equity stake. The company’s emergence aligns with Google X’s recent strategy of aggressively spinning off projects following workforce reductions.

Link to full article