Anthropic has expanded its Claude Cowork feature to support remote session management from mobile devices, allowing users to initiate, monitor, and control complex AI-driven workflows across smartphones, tablets, and desktop environments. The update, currently rolling out in beta to Max plan subscribers, transforms Claude Cowork from a desktop-bound productivity tool into a persistent, cross-platform AI agent capable of executing multi-step tasks even when the user’s primary device is offline.
Claude Cowork functions as an AI agent designed to handle entire workflows rather than generating single responses. It integrates with email, calendars, messaging platforms, local files, and web resources to automate business processes that traditionally required hours of manual effort. According to usage data shared by Anthropic, over 90 percent of Cowork activity involves non-programming tasks, with the majority concentrated in business operations and content creation.
How Claude Cowork’s Remote Session Management Works
The core capability of the update is cross-device session continuity. A user can initiate a task on a laptop, such as analyzing financial data or compiling a client presentation, and then monitor progress, review intermediate outputs, or provide input from a smartphone without interrupting the workflow. Scheduled tasks can run autonomously on a predetermined timeline. For example, a user can assign Claude to analyze overnight emails, meeting transcripts, and news updates, then generate a briefing document before the start of the workday.
Critically, Claude Cowork introduces real-time decision checkpoints. When the AI encounters a step that requires human judgment, it sends a prompt directly to the user’s mobile device. No final output is delivered or shared without explicit user approval, maintaining a human-in-the-loop safeguard that prevents the AI from acting autonomously on sensitive decisions.
This architectural choice reflects a deliberate balance between automation and control. The system can continue processing in the background even when the user’s device goes offline, executing scheduled tasks and pausing only when human input is needed.
Security Implications of Persistent, Cross-Platform AI Sessions
From a cybersecurity perspective, the expansion of Claude Cowork to mobile devices and persistent background execution introduces several material risk considerations that security teams should evaluate before permitting enterprise use.
Cross-device synchronization and continuous background processing increase the attack surface. Mobile devices, which often lack the same security controls as managed desktop environments, become potential entry points for unauthorized access to Cowork sessions. If a smartphone is compromised through malware, phishing, or device theft, an attacker could potentially intercept session tokens, view task progress, or access enterprise data that the AI agent has retrieved.
Claude Cowork’s integrations with multiple third-party tools create additional vectors for data leakage. The agent accesses emails, calendars, messaging applications, local file systems, and web-based resources to execute workflows. Each integration represents a potential lateral movement path if authentication mechanisms are weak or session isolation is insufficient.
Session Management and Access Control Risks
The consolidation of chat interactions and Cowork sessions into a unified interface raises questions about session management and access control. If authentication mechanisms rely solely on device-level security rather than strong, context-aware verification, threat actors could potentially hijack active workflows. Security professionals should examine how Claude handles encryption of session tokens, whether session isolation is enforced between different users on shared devices, and what device-level authentication requirements are mandated.
Anthropic has emphasized that users retain final approval authority over all outputs, which reduces the risk of unintended data exposure from autonomous AI actions. However, this safeguard does not address scenarios where an attacker gains legitimate access to an active session through credential theft or session hijacking.
Enterprise Adoption Considerations for AI Agents
Organizations evaluating Claude Cowork for workflow automation should implement a layered security approach that addresses the specific risks of persistent AI agents operating across environments. Access controls should follow the principle of least privilege, restricting which users can create, modify, or monitor Cowork sessions. Activity logging should capture session creation, data access events, decision checkpoint responses, and output delivery to enable forensic analysis if an incident occurs.
Mobile device security policies become critical when Cowork sessions can be managed from smartphones. Organizations should enforce device-level encryption, require biometric or hardware-backed authentication, and consider deploying mobile endpoint detection and response solutions for devices that access enterprise AI workflows. A multi-layer endpoint protection solution with real-time threat detection and behavioral analysis is appropriate for any device that interacts with persistent AI sessions handling sensitive business data.
What Organizations Should Do Now
The productivity gains from persistent, cross-platform AI agents are significant, but the security implications require deliberate management. As a first step, security teams should audit which users and devices currently have access to Claude Cowork, verify that session timeout policies are configured appropriately, and ensure that mobile devices used for remote session management comply with organizational security standards. Implementing strict access controls and monitoring activity logs for anomalous session behavior will help mitigate the elevated risk profile that accompanies this new class of autonomous, cross-platform AI operations.
