Anthropic bought $3.4 billion worth of natural gas turbines in a push to secure power for its data centre operations. The purchase marks the AI safety firm's entry into a broader industry trend of major AI companies investing directly in energy infrastructure to fuel their expanding computational demands. Source
Meta released Muse Ember, a lightweight model designed for agent execution, priced at 10 cents per million input tokens—significantly cheaper than frontier competitors. The release extends Meta's portfolio of increasingly cost-conscious models aimed at developers building agent-based applications at scale. Source
Anthropic announced Claude Fable 5.2 and Claude Mythos 5.2, new models the company says show improvements in coding, reasoning, and vision while generating outputs faster than their 5.1 predecessors. The release comes as Anthropic prepares for its reported initial public offering, with the company expected to set its valuation range this week. Source
Researchers at Google DeepMind used an AI tool to formally verify that the Signal Protocol, a widely used encryption standard, works as intended. The team's formal verification achievement represents a step toward their broader goal of applying the same techniques to verify "all the software of the world." Source
Claude Code is exposing thousands of secrets on GitHub after a "do not ask again" prompt in settings bypasses safety measures, allowing passwords, API keys, and other sensitive credentials to be shared in code repositories. The exposure has affected multiple users relying on Anthropic's coding assistant feature. Source