Agents gain a browser, Apple sues OpenAI, and Cloudflare reroutes caching

Anthropic gave Claude Code on desktop an in-app browser, a survey finds 57% of enterprises have seen AI agents be confidently wrong, and Apple has sued OpenAI and some of its employees over alleged trade secret theft.

What agents can see, and what they get wrong

Claude Code on desktop can now open documents, designs and websites in a built-in browser, so developers can pull reference material into a session without switching to another app. The feature is about gathering context in one place, and that is the problem a VentureBeat report raises from the enterprise side. It says 57% of enterprises have watched AI agents make confidently incorrect decisions and argues that an agentic context layer could reduce those errors. The same report also asks who actually has one, which suggests the remedy is still more proposal than standard practice.

Apple's trade secret case against OpenAI

Apple has filed suit against OpenAI and individuals it names, alleging that confidential trade secrets and information about unreleased hardware reached OpenAI through former Apple employees. These are allegations in a complaint, and no court has ruled on them. For engineers who move between large companies, the case is a reminder that what former employees carry with them about unreleased hardware can become the subject of litigation, though its outcome and any wider effect on hiring remain open.

Keeping cache traffic on the right continent

Cloudflare has improved Smart Tiered Cache for public cloud regions, targeting latency that comes from requests being routed across continents when they do not need to be. The company says the change should help applications whose origins sit in public cloud regions, because cache fills can stay closer to where the origin actually runs. Teams serving traffic from a single cloud region are the most likely to notice the difference, though Cloudflare's post is the only source for the expected gains.