The Information

Anthropic hiring custom AI chip design team

Anthropic is assembling an internal “Custom Silicon Team” of hardware and software experts to co-design custom AI chips and models intended to make its Claude models faster and more efficient, confirming earlier reports. The company has held talks with Samsung to manufacture a chip it designs and is reported to be collaborating with Broadcom and Google on specialized Tensor Processing Units. Anthropic still relies on infrastructure and hardware relationships with Amazon Web Services, Google Cloud, Nvidia and AMD, but sees in-house silicon as necessary to scale. Industry estimates cited by Reuters put the cost of designing a modern AI chip at roughly $500 million, and timing for a first Anthropic-designed chip remains unknown. Google has also agreed to invest in Anthropic under a multibillion-dollar arrangement announced in April.

OpenAI launches ChatGPT Work agent

OpenAI released ChatGPT Work on July 9, 2026 — an agent product for knowledge work that consolidates prior agent efforts (Codex, ChatGPT agents, OpenClaw) into a cloud-hosted workspace. Work runs on the Codex harness, uses an isolated cloud microVM and a managed Chrome browser service, integrates with plugins/connectors to services like Gmail, Slack, Salesforce, Notion and Figma, and can produce interactive artifacts (sheets, docs, slides, Sites). Work supports cloud and local desktop modes, scheduled/automated tasks, and per-task persistent working directories; three weeks after launch Work (and Codex) reportedly crossed 10 million users. OpenAI plans to merge Chat and Work modes into a single experience by the end of the year.

White House Hosts AI Firms to Review Model-Testing Framework

The White House will meet with leading artificial intelligence companies to review a newly completed voluntary framework for testing the cybersecurity capabilities of advanced AI models. Representatives from Anthropic are expected to participate, and OpenAI and Google are also reported to attend. The voluntary program would let participating developers give the government early access to certain frontier models for up to 30 days; the framework cannot be used to create mandatory federal licensing, permitting or preclearance requirements. The meeting follows a June 2 executive order directing federal agencies to create a process to identify “covered frontier models” and to develop classified benchmarks for assessing advanced cyber capabilities.