The afternoon briefing.

Frontier AI Models. Anthropic has publicly released Claude Fable 5, a "Mythos-class" model previously deemed too powerful for general access. This new model demonstrates significant advancements in coding and scientific tasks, with one instance completing a Stripe code migration in a single day that would typically take two months. Google also advanced its AI capabilities with Gemini 3.5 Live Translate, offering real-time voice translation across more than 70 languages while preserving speaker nuances. These releases signal a rapid acceleration in accessible, powerful AI tools.

AI Regulation & Policy. Apple announced that its new Siri AI features will not launch in the European Union, directly attributing the delay to the Digital Markets Act (DMA). This move highlights the ongoing tension between tech giants and regulatory bodies, with Apple suggesting the DMA's interoperability requirements pose security and privacy risks. Separately, Meta quickly pulled facial recognition code from its smart glasses, just one day after its discovery, indicating a cautious approach to privacy in hardware development. These incidents underscore the increasing impact of policy on AI deployment and feature availability.

AI Infrastructure & Economics. China has activated the world's first underwater data center, a novel approach to managing the growing energy demands of AI and computing infrastructure. This development, alongside NVIDIA's new DGX Spark Enterprise Manageability, reflects the industry's focus on scalable and efficient AI infrastructure. Simultaneously, discussions are emerging around the potential for tech companies to adopt cheaper AI models, which could significantly alter the economic landscape of AI development and deployment. These efforts aim to optimize the foundational elements supporting the AI boom.

AI Security & Privacy. The Java ecosystem faces a new security emergency, with AI tools now capable of exploiting vulnerabilities in older frameworks like Spring. This development highlights the evolving threat landscape where AI itself can be a vector for security risks. In a related move, the US CISA has issued an urgent directive for federal agencies to patch a critical VPN bug actively exploited by a ransomware group, emphasizing the immediate need for robust cybersecurity. These events underscore the critical importance of integrating security measures into AI development and infrastructure.