EU AI Act enforcement begins as Google's AI Co-Scientist demonstrates reasoning breakthrough in antibiotic research
The European Commission began enforcing the AI Act on 2 August 2026, while Google's AI Co-Scientist solved a decade-long antibiotic resistance puzzle in weeks.
EU AI Act enforcement begins with new transparency rules
On 2 August 2026, the European Commission’s AI Office, together with national authorities, began enforcing the Artificial Intelligence Act. As part of this enforcement, new transparency rules under the EU AI Act started to apply the same day, marking a significant shift in how AI systems must operate across Europe.
Google’s AI Co-Scientist solves antibiotic resistance puzzle
In a striking demonstration of AI reasoning capabilities, Google announced its AI Co-Scientist in May 2026. The system was given a one-page brief with the goal of figuring out how antibiotic resistance spreads between bacterial species.
AI Co-Scientist concluded that antibiotic resistance genes were spreading between bacterial species by hitching rides on bacterial viruses—a finding that took on particular significance when researchers at Imperial College London revealed they had spent a decade reaching the same conclusion through wet-lab work, with their paper still in peer review at the time.
Reasoning, not just data, drives scientific AI
The breakthrough reflects arguments made by Eric Schmidt and Suhas Mahesh in their article ‘AI for science needs reasoning, not just data,’ published in MIT Technology Review. Schmidt, who served as CEO of Google from 2001 to 2011, has since shifted focus to scientific innovation. In 2024, Schmidt and his wife Wendy co-founded Schmidt Sciences, a philanthropic venture to fund unconventional areas of exploration in science and technology.
The significance of such advances becomes clear when considering the field’s historical importance: over 25 Nobel Prizes have relied on protein crystallography. AI systems like AlphaFold have already demonstrated the potential to transform this field by predicting the three-dimensional structures of proteins by learning from thousands of experimentally measured shapes.
Source: MIT Technology Review / European Commission Press Corner