Cyber Security News

Claude Cowork Expands to Web and Mobile With Background AI Agent Workflows

Claude Cowork was expanded beyond the desktop, rolling out web and mobile versions that let AI-driven task sessions persist across devices and continue running in the background without an active connection.

The beta rollout begins with Max-tier subscribers over the coming weeks, with additional plan tiers to follow.

Cowork operates as an agentic workspace where users delegate multi-step tasks to Claude, which then works across connected files, calendars, email, messaging apps, and web resources until completion.

According to Anthropic’s usage analysis, more than 90% of Cowork activity involves non-coding tasks, contrasting with the coding-heavy reputation most AI agents have built.

The largest use categories are business operations and content creation, including reconciling quarterly spend into variance memos, converting contract folders into renewals trackers with flagged risks, and building client presentations from call transcripts and pipeline data.

Claude Cowork Expands to Web and Mobile

Three architectural shifts define this release:

  • Cross-device continuity: Sessions and files now sync across desktop, web, and mobile, allowing users to initiate a task on one device and monitor or resume it on another.
  • Offline background execution: Scheduled tasks can now run without any device online. A task queued for 6 a.m., for example, can process email threads and recent news, compile a briefing document, and draft a follow-up email for review, entirely unattended.
  • Human-in-the-loop checkpoints: When Claude encounters a decision requiring human judgment, it pauses and sends the query to the user’s phone. Nothing is finalized or sent until explicit approval is given.

Anthropic emphasizes that desktop remains the most capable environment, since only the desktop client can access local files and the browser directly. The mobile and web versions extend reach to users who previously couldn’t install a desktop application.

Claude Cowork goes cross-device with web, mobile, and offline AI task execution.

On web and desktop, chat and Cowork sessions now share a unified interface, with projects and artifacts synchronized across both surfaces. This consolidation reduces context-switching for users managing multiple concurrent agent tasks.

For security teams and content operations relying on AI agents, the shift to persistent, background-executing, cross-device sessions raises practical considerations around credential exposure, task audit trails, and approval-gate integrity, particularly since Cowork integrates with email and messaging apps that may carry sensitive organizational data.

The “human approves before anything ships” model addresses some autonomous-action risk, but organizations adopting Cowork at scale should evaluate logging and permission scoping for connected tools.

Web access launches directly from the claude.ai home screen. Mobile access is available via the Cowork option in the sidebar of the Claude iOS and Android apps. The full-featured experience, including local file and browser access, remains exclusive to the Claude desktop app.

Anthropic has also extended its doubled Cowork usage limits promotion through August 5, 2026, encouraging users to test larger, more complex delegated tasks during the transition period.

Kavichselvan

Recent Posts

Hackers Use ConnectWise SkyKick Credentials to Access Microsoft 365, Halting Cloud Backups

Attackers abused compromised ConnectWise Cloud Backup credentials to access Microsoft 365 customer environments, prompting ConnectWise…

1 day ago

Anthropic’s Claude AI Uncovers Origins of NSA-Provided Cryptographic Keys After 25 Years

Anthropic says an internal Claude research model reconstructed how the National Security Agency generated seeds…

2 days ago

Critical WordPress Flaws Allow Attackers to Execute Remote Code Without Authentication

Two WordPress Core vulnerabilities can be chained to execute remote code without authentication, according to…

2 days ago

OpenBSD Kernel Flaw Dating Back 37 Years Enables Local Privilege Escalation to Root

A newly disclosed OpenBSD kernel exploitation chain combines a shared memory accounting flaw with a…

2 days ago

GhostAction Hackers Hide Credential-Stealing Malware in Fake Font Files Across 94 GitHub Repositories

GhostAction attackers have expanded their GitHub credential theft campaign with a second stage that disguises…

2 days ago

SideCopy Hackers Target Indian Organizations With ReverseRAT Malware to Steal Data

Indian authorities have warned that SideCopy hackers are targeting Indian organizations, including academic institutions, with…

2 days ago