Actus et Outils sur l’IA | Intelligence Artificielle

DeepSeek V4 Pro : le champion du prix cassé change de stratégie

19 août 2026 at 12:00

DeepSeek V4 Pro: Price War Takes a New Turn

DeepSeek's V4 Pro launch is followed by a sharp price hike, Google answers with a cheaper Gemini 3.7 Flash, and OpenAI completes a $7 billion employee tender offer.

Twitch entraîne l'IA d'Amazon avec vos streams : comment refuser

DRAMA IA

17 août 2026 at 10:00

Twitch AI Training on Your Streams: How to Opt Out

Twitch quietly enabled AI training on your streams, VODs, and chat by default. Here's where the opt-out setting is hidden and how to switch it off.

IA

17 août 2026 at 08:00

AI Infrastructure: Why the Billions Are Changing Sides

Nvidia mobilizes $500 billion for AI infrastructure, Anthropic locks in a 20-year power deal, and Google reshuffles DeepMind — three signals of the same shift.

Quarante minutes en ligne, 2 500 entreprises exposées

IA

15 août 2026 at 17:00

LiteLLM Supply Chain Flaw: 2,500 Companies Exposed

Two trapped versions of LiteLLM stayed on PyPI for about 40 minutes. CloudSEK reports 2,500 companies and 434,000 CI/CD pipelines potentially exposed by the supply chain attack.

Claude va filigraner ses textes, et c'est l'Europe qui l'a décidé

IA

14 août 2026 at 20:00

The AI Act Made Anthropic Watermark All of Claude’s Text

Anthropic will soon mark every text Claude generates, even outside Europe. The trigger: the EU's AI Act, in effect since August 2, 2026, and the backlash has already started.

GPT-5.6-Cyber : OpenAI livre un modèle qui trouve des failles

IA

13 août 2026 at 10:00

GPT-5.6-Cyber: OpenAI’s Model That Finds Flaws

OpenAI launched GPT-5.6-Cyber, a cybersecurity-focused model that accepts 95% of advanced security requests and found two new flaws in Chrome. Access remains tightly restricted.