Anthropic ujawnia 133 mln czatów z wyłączonymi filtrami biobroni

Luka w zabezpieczeniach Anthropica, rekordowe modele open-source i nowy benchmark percepcji wizualnej AI.

Źródła:

• Anthropic ran 133 million contractor chats with its bioweapon filters off
• Alibaba's Qwen team releases Qwen 3.8 models with open weights under the Apache 2.0 license
• Study contradicts Anthropic and OpenAI claims that autonomous AI research is within reach
• Zhipu AI releases GLM-5.3, claims it's the strongest open-weights coding model
• GPT-5.6 Sol goes 14x faster as OpenAI launches Ultrafast mode powered by Cerebras
• New benchmark confirms AI models still perform poorly at visual perception
• 15% of AI page fetchers in Europe reached disallowed URLs, TollBit finds
• The AI store manager fired its first human. It had to be reminded of its own rules first
• OpenAI’s new ChatGPT feature logs your keystrokes and stores them in plain text
• Hey, Your AI Is Showing
• How Claude's text watermarking works \ Anthropic
• The "tragedy of the cognitive commons" explains how rational AI adoption could destroy entire professions' expertise
• Etzioni on AI: Become a power user

📰 Przeczytaj całe wydanie na stronie

🤖 Odcinek opracowany przy pomocy sztucznej inteligencji. Lektor syntetyczny.