If you think the current AI boom is just about smarter chatbots, you’re missing the shift from "thinking" to "doing"—and the chaos that comes with it. The real story here isn't just that your AI is getting a mobile app; it’s that the tech industry is sprinting toward "agentic" systems that operate in the background of your life, often while you aren't even watching.
This week, Anthropic announced a major expansion of its Claude Cowork agent, which is moving to the cloud to allow users to manage tasks across mobile and web interfaces, according to NBC News. Previously, Cowork was tethered to a single desktop, requiring that machine to stay powered on and connected to finish a task. Now, as TechCrunch notes, the system can run independently of your device's status, essentially acting as a digital administrative assistant that works while you sleep. **Engadget]() clarifies that while this doesn’t mean the AI is automating your phone's apps, it provides a crucial "command center" to monitor and approve tasks being performed on your desktop back home.
But beneath this productivity polish lies a volatile reality. While Anthropic markets these tools to non-technical users, the company is simultaneously embroiled in a high-stakes geopolitical tug-of-war. **CBS News]() reports that the federal government recently lifted export controls that had effectively blocked Anthropic from deploying its powerful Fable 5 and Mythos 5 models. This followed a weekslong standoff where the Trump administration had restricted access due to concerns over "jailbreaks"—a claim Anthropic publicly contested, arguing the recall standard would halt the entire industry’s progress.
The friction doesn't stop at the U.S. border. As Anthropic pushes to dominate the "agentic" office space, it has been aggressively monitoring users to prevent what it calls "distillation attacks," where rival labs—specifically those in China—attempt to clone the capabilities of its models. Ars Technica reveals that a security researcher discovered "prompt steganography" hidden within Claude Code that tracked user data, including timezones and potential connections to Chinese AI labs. The backlash was immediate: Alibaba has since banned its employees from using Claude Code, labeling it a "high-risk software" with security vulnerabilities, according to Ars Technica.
This paints a picture of a company trapped between a mandate for hyper-utility and a necessity for surveillance. Anthropic claims the tracker was an "experiment" to protect intellectual property, but the move has alienated parts of its user base who view such monitoring as a betrayal of the company’s stated anti-surveillance stance. It is a classic Silicon Valley paradox: building tools designed to be "invisible" and helpful, while simultaneously becoming a deeply intrusive gatekeeper.
For the average user, the takeaway is clear: the convenience of having an AI draft your emails while you're offline comes with an invisible trade-off. Anthropic is currently extending its doubled usage limits for Cowork through Aug. 5, as reported by NBC News. Watch how the company navigates its next federal hurdle; the government's ongoing appeal against a court ruling that previously blocked restrictions on Anthropic’s military contracts suggests the peace between the White House and the AI firm is, at best, a fragile ceasefire.
