Qwen3.8 is launching and will be released as an open-weight model soon.

Researchers demonstrated a method to deceive GitHub's AI agent, causing it to leak contents from private repositories. The findings are detailed in a blog post by Noma Security.

LongCat-2.0是一款大规模**Mixture of Experts (MoE)**模型，总参数量达1.6万亿，活跃参数量为480亿。公告中包含博客链接及Hacker News讨论帖链接。

据《金融时报》2026年6月28日报道，谷歌已限制Meta使用其Gemini AI模型。

Asian AI startups are launching new models with Mythos-like capabilities, potentially filling a market gap due to an export ban on similar models from U.S. AI labs, such as Anthropic.

DeepSeek has open-sourced inference optimizations, resulting in 60–85% faster generation, as detailed in the DSpark paper available on GitHub.

人工智能公司 Anthropic PBC 指控阿里巴巴集团有限公司（Alibaba Group Holding Ltd.）使用数千个虚假账户非法访问其 Claude AI 模型，违反了中国地区产品可用性限制。

OpenAI announced that the U.S. government will determine which users can access its latest AI model upgrade. The decision follows a report by The Washington Post.

Anthropic afferma che Alibaba ha illicitamente estratto le capacità del modello AI Claude. L'articolo è pubblicato da Reuters il 24 giugno 2026.

Anthropic hat ein neues Tool namens 'Claude Tag' vorgestellt. Der Artikel enthält einen Link zur offiziellen Ankündigung sowie eine Diskussion auf Hacker News mit 72 Punkten und 33 Kommentaren.