The Blog

OpenAI says its forthcoming Astra model is the first to cross its "critical" cybersecurity capability threshold, able to find and exploit unknown flaws without human guidance. Nvidia is reportedly closing in on a roughly $14 billion deal to buy Hugging Face. And PwC forecasts global AI infrastructure investment will reach $31.6 trillion through 2050.

Anthropic released Claude Opus 5 at half the cost of Claude Fable 5, with Math Olympiad-level reasoning. Security researchers at Anthropic and EPFL discovered self-propagating payloads can spread between AI agents. And Meta shipped Llama 4 with a 10-million-token context window and native multimodality.

A critical Screen Sharing flaw is letting attackers gain root access on internet-exposed Macs and plant cryptocurrency miners. xAI ships Grok 4.6, matching GPT-5.6 Sol on independent benchmarks. And UBS estimates that OpenAI and Anthropic alone could drive 48% of Google Cloud's revenue next year.

An unpatched GeoServer SQL injection flaw is already being probed for remote code execution, hours after disclosure. DeepSeek ships V4 Pro, its flagship model with a 1 million token context window. And Amazon raises its 2026 AI infrastructure spending to around $220 billion, saying capacity will stay tight through 2027.

A compromised build pipeline in the widely used LiteLLM project exposed more than 434,000 CI/CD pipelines across 2,500+ companies. Alibaba launches Qwen3.8-Max, a 2.4-trillion-parameter open-weight model aimed at Anthropic and OpenAI. And Microsoft's Azure cloud business crosses $100 billion in annual revenue for the first time.

Microsoft's August update fixes 394 vulnerabilities, including three zero-days with one already under active attack. Anthropic starts adding invisible watermarks and provenance metadata to everything Claude generates, worldwide. And Nvidia lines up six Wall Street firms to mobilise over $500 billion in financing for AI data centres and compute.