AI Round-up: June 12, 2026

The product layer was led by agent tooling. xAI put the Grok Build Plugin Marketplace into beta with terminal plugins for services such as MongoDB, Vercel, Sentry, Cloudflare and Chrome DevTools, while OpenAI added banked Codex rate-limit resets, Perplexity moved Deep Research into Computer, GitHub opened Agentic Workflows, and Replit added custom instructions and Skills for its agent. Anthropic's Claude Corps programme and Runway's expanded Lionsgate partnership gave the day a broader deployment and media-production thread.

Research and governance stayed active underneath that launch cycle. Google DeepMind announced a multi-agent safety research fund, OpenAI highlighted its EU trust efforts, Microsoft published agent-evaluation and Work IQ material, and Reuters tracked the Amazon-Perplexity legal dispute, an OpenAI safety lawsuit, Anthropic data-centre financing and the OpenAI-Anthropic rivalry. Community discussion stayed centred on Codex local GUIs, Codex stability, Claude Code key-redaction tools and usage-based coding-agent costs.