Morning Edition · 22 Iulie 2026
23 JULY
Main story
Edition stories
EVALUATION 2026: To get a good grade, they stole the answers and the grading key during the exam itself
A cybersecurity experiment involving advanced AI models revealed unexpected autonomous behavior: instead of solving the ExploitGym benchmark tasks, an AI agent identified a vulnerability …
Did We Measure AI’s Environmental Impact Incorrectly? A UN Study Explains Why a Single Metric Is No Longer Enough
A report published in June 2026 by the United Nations University Institute for Water, Environment and Health (UNU-INWEH), The Environmental Footprint of AI, argues …
Google Bets on Speed, Not Power. Three New Gemini Models, but the Question Remains: Where Is Gemini 3.5 Pro?
Google announced, in a post on its official blog, the launch on Tuesday, July 21, 2026, of three new models in the Gemini family: …
China Begins to Close the Doors Once Wide Open to the AI Community
China’s Ministry of Commerce has begun consultations with major domestic artificial intelligence and semiconductor companies, including Alibaba, ByteDance and Zhipu, to find ways of …
Airbus Is Moving Its Critical Applications Away from Amazon Web Services. Why Sovereign Cloud Is Becoming a Strategic Priority in Europe
Airbus announced on Thursday, July 16, 2026, a multi-year agreement with Scaleway, the cloud computing subsidiary of France's Iliad Group, under which approximately 70 …
AI With an Accent: How Claude Changes Its "Personality" Depending on the Language You Speak
The language used in a prompt may itself become a control variable for shaping AI responses. A new Anthropic study suggests that language choice …
Autopsy of a Foretold Failure: What Sora's Collapse Teaches Us
Sora's disappearance is more than the story of a product being withdrawn from the market—it is one of the first clear signals that, in …
Google Makes Its Biosecurity Strategy Public: The Same AI Models That Could Stop a Pandemic Could, in Theory, Also Help Cause One
Major technology companies are beginning to confront an uncomfortable question: what happens when technology can accelerate scientific discoveries with direct implications for biosecurity?
When an AI Assistant Knows Too Much. Grok Build Was Sending Entire Projects to xAI's Servers Without Users' Knowledge
Grok Build, xAI's AI coding assistant, was transmitting not only the files required to generate responses but users' entire software projects to the company's …
RO AI Factory – Romania's First AI Factory, Built Through the European EuroHPC Initiative
Romania is joining the European network of AI Factories through a project that has the potential to transform how artificial intelligence applications are developed …
The Myth of Creativity’s Immunity to Automation
How the Hollywood Strikes Became the First Major Labor Conflict Triggered by Generative Artificial Intelligence
UiPath Foundation Launches a Free Artificial Intelligence Course for High School Students
UiPath Foundation, the non-profit organization of the Romanian company UiPath, has created AI Generation — a free course through which high school students (grades …
EU Partially Delays Enforcement of Key AI Act Rules — but Bans “Nudify” Applications
The European Union has decided to postpone the enforcement of several key provisions of its flagship artificial intelligence law, giving technology companies up to …
U.S. Government Allows Limited Resumption of Access to Claude Mythos 5
Anthropic announced on Friday, June 26, 2026, that U.S. authorities had allowed access to its advanced cybersecurity model, Claude Mythos 5, to resume for …
Intelligence — Now Artificial; Wisdom — Always Human
Intelligence — Now Artificial; Wisdom — Always Human
How AI changes administrative work
Document automation can reduce lost time, but it requires well-defined processes and responsibilities.