AI IN LIFE – AI Newshttps://ai-in-life.com/en/The most important AI news of the day, daily – by AI IN LIFE.enEU AI Act: Transparency duties and GPAI rules in force since 2 Augusthttps://ai-in-life.com/en/news/eu-ai-act-transparenzpflicht-august-2026/https://ai-in-life.com/en/news/eu-ai-act-transparenzpflicht-august-2026/Chatbots must reveal they are AI, deepfakes and AI content must be labeled – and providers of large AI models now fall under the direct oversight of the European Commission.REGULATION2026-08-08AI designs first working viruses – 16 bacteriophages built by computerhttps://ai-in-life.com/en/news/ki-designt-erste-funktionierende-viren-stanford-arc-institute/https://ai-in-life.com/en/news/ki-designt-erste-funktionierende-viren-stanford-arc-institute/Researchers at Stanford and the Arc Institute used AI models to design complete viral genomes. Sixteen of them worked in the lab and killed E. coli bacteria – a scientific breakthrough with serious governance questions.RESEARCH2026-08-08OpenAI slows new model 'Astra' – first-ever 'critical' cyber capabilitieshttps://ai-in-life.com/en/news/openai-pausiert-astra-kritische-cyber-faehigkeiten/https://ai-in-life.com/en/news/openai-pausiert-astra-kritische-cyber-faehigkeiten/OpenAI has slowed development of its frontier model Astra because it could reach the highest risk tier for offensive cybersecurity. It is the first time the company's internal safety framework has crossed that threshold.SAFETY2026-08-08Anthropic: AI classifier replaces human approval in Claude Codehttps://ai-in-life.com/en/news/anthropic-claude-code-ki-classifier-ersetzt-freigabe/https://ai-in-life.com/en/news/anthropic-claude-code-ki-classifier-ersetzt-freigabe/From 14 August, auto mode becomes the default: instead of humans signing off on risky commands, an AI classifier reviews them. The reason is uncomfortable – humans caught dangerous commands only 13.6% of the time.INDUSTRY2026-08-07ByteDance is training an AI model with up to 10 trillion parametershttps://ai-in-life.com/en/news/bytedance-trainiert-10-billionen-parameter-modell/https://ai-in-life.com/en/news/bytedance-trainiert-10-billionen-parameter-modell/The TikTok owner's Seed division is working on a model of unprecedented size for a Chinese company, according to the Financial Times – a signal in the race for sheer scale.CHINA2026-08-07xAI ships Grok Imagine Image 2.0 with precise image editinghttps://ai-in-life.com/en/news/xai-grok-imagine-image-2-0-bildbearbeitung/https://ai-in-life.com/en/news/xai-grok-imagine-image-2-0-bildbearbeitung/The new image model from Elon Musk's xAI enables targeted partial edits and up to five reference images at once – and, by its own account, ranks second on global arena leaderboards.PRODUCT2026-08-09