Anthropic AI Updates: August 6, 2026

1. Anthropic Builds Custom AI Chip Design Team

Anthropic. Anthropic is assembling a custom silicon team to design its own AI chips, saying it will co-design hardware and models to help its technology run faster and more efficiently. The effort aims to reduce reliance on infrastructure partners such as AWS, Google, Nvidia, and AMD as demand for Claude grows, and joins similar in-house chip programs at OpenAI, Google DeepMind, and Meta. Source

2. Inference Hooks Add Inline Data Loss Prevention to Claude Enterprise

Anthropic. Anthropic introduced inference hooks, a data loss prevention feature for Claude Enterprise that routes each prompt and tool call through an organization’s security server over a signed WebSocket connection before Claude processes them. The feature covers all Enterprise surfaces including chat, Claude Code, and Claude Cowork, integrates with DLP platforms such as Netskope, Palo Alto Networks, and Proofpoint, and supports staged rollouts through shadow mode and percentage-based testing. Source

3. Anthropic Details Cost Visibility and Control Tools for Claude Enterprise

Anthropic. Anthropic published a guide outlining how administrators can monitor and manage AI spending across Claude Enterprise using access gating, model controls, and hard spend caps. The post also covers a usage analytics dashboard, an Analytics API, and API-level optimizations such as prompt caching, batch processing, and the effort parameter. Source

4. Anthropic Signs $10 Billion Cloud Deal With Startup Volta

Anthropic. Anthropic signed a $10 billion, six-year agreement with Volta, a newly founded AI cloud provider that partners with crypto-mining firm Bitdeer to build a 133-megawatt data center in Norway powered by Nvidia’s Vera Rubin chips. The deal adds to Anthropic’s rapid infrastructure buildout alongside recent SpaceX and Amazon partnerships as frontier labs race to secure compute capacity. Source