The AI week in 5 points
The essentials of this week's AI news, summarized for those who don't have time to follow it all.
No time to follow AI news day by day? Here are the 5 things to remember from this week.
1. Europe moves to label AI content
From August 2, the AI Act requires clearly flagging AI-generated content (deepfakes, text, chatbots). Transparency becomes a legal obligation.
2. AI safety goes operational
More than 1,100 employees of OpenAI, Anthropic, Google and Meta signed a letter calling for a coordinated, verifiable “pacing” mechanism for the most advanced AI.
3. Agent security attracts capital
Nearly $855M raised in 2026 to secure AI agents, plus an acquisition announced around one billion dollars. A new front to watch.
4. Open-source keeps closing the gap
Open models increasingly rival proprietary ones, paving the way for sovereign enterprise deployments.
5. The assistant race tightens
ChatGPT, Gemini and Claude have never been closer in quality: the choice now comes down to price, ecosystem and use case. (More in our comparison.)