Claude Cowork Critical Flaw Exposes AI Sessions to Remote Access

Key Takeaways

• Claude has rolled out a significant update to its Cowork feature, allowing remote management and continuation of AI-driven sessions across mobile and web platforms.
• This update enables persistent AI task execution, delegating complex workflows to Claude and monitoring progress from any device.
• While enhancing productivity, the expanded cross-device access and persistent background execution increase the attack surface, raising cybersecurity concerns for sensitive enterprise data.
• The feature is currently in beta, launching first for Max plan subscribers.

Anthropic’s Claude has introduced a substantial enhancement to its Cowork functionality, enabling users to remotely control and resume AI-powered sessions across both mobile and web interfaces. This update represents a strategic pivot towards continuous AI task execution, empowering users to assign intricate workflows to Claude and track their advancement from any device, including smartphones.

Previously confined to desktop environments, Claude Cowork now offers seamless cross-device continuity. This means a user can initiate a task on a laptop and effortlessly pick up or review its progress from a mobile device without interruption. The feature is currently in its beta phase and is being progressively rolled out, commencing with users subscribed to the Max plan.

Claude Cowork functions as an AI agent engineered to manage multi-step tasks across various integrated tools, such as email, calendars, messaging applications, local files, and web resources.

Claude Cowork Controls AI Sessions

Unlike traditional AI chat interactions, Cowork’s primary focus is on executing complete workflows rather than delivering single responses. Data shared by the company indicates that over 90 percent of Cowork activities are dedicated to non-programming tasks, predominantly involving business operations and content creation.

This trend underscores the increasing integration of AI into routine enterprise workflows. Typical applications include analyzing financial data, generating reports, organizing contract repositories, and preparing client presentations drawing from multiple data sources. Such tasks frequently span several hours or even days, making continuous background processing a vital capability. A key enhancement is Claude’s capacity to continue working even when the user’s device is offline. Scheduled tasks can operate autonomously, allowing users to pre-assign jobs like preparing meeting briefs or drafting communications.

For instance, a user could schedule Claude to analyze emails, transcripts, and news updates early in the morning, generating a comprehensive briefing document before the start of the workday. The system also incorporates real-time decision checkpoints. Should Claude encounter a step requiring human intervention, it dispatches a prompt directly to the user’s phone. This mechanism ensures users retain control over critical decisions while the AI manages execution. Crucially, no final output is delivered or shared without explicit user approval, maintaining a human-in-the-loop safety measure.

Cybersecurity Implications

From a cybersecurity standpoint, this expanded access model introduces new considerations. Cross-device synchronization and persistent background execution enlarge the potential attack surface, particularly if mobile devices become compromised. Unauthorized access to Cowork sessions could expose sensitive enterprise data, including emails, internal documents, and business analytics. Furthermore, integrations with numerous third-party tools might create additional entry points for data leakage if not adequately secured.

The consolidation of chat and Cowork into a unified interface also raises concerns regarding session management and access control. If authentication mechanisms are weak or sessions are not properly isolated, threat actors could potentially hijack active workflows. Security professionals may need to assess Claude’s handling of encryption, session tokens, and device-level authentication to mitigate these risks. Claude has emphasized that users maintain final approval over all outputs, which helps reduce the likelihood of unintended data exposure.

However, organizations adopting AI agents for workflow automation should implement stringent access controls, monitor activity logs, and enforce robust mobile device security policies. This update reflects a broader industry movement toward autonomous AI agents capable of executing complex tasks across diverse environments. While the productivity gains are substantial, the security implications of persistent, cross-platform AI operations will necessitate careful scrutiny as adoption rates climb.

What You Should Do

• Implement strong, unique passwords and multi-factor authentication (MFA) for all accounts linked to Claude Cowork.
• Ensure all mobile devices used to access Cowork sessions are secured with up-to-date operating systems, security patches, and endpoint protection.
• Regularly review access permissions for Claude Cowork and any integrated third-party applications to adhere to the principle of least privilege.
• Monitor activity logs for unusual or unauthorized access patterns within Cowork sessions.
• Educate users on the risks associated with AI agent access and the importance of secure device usage and prompt approval processes.

Disclaimer: HackersRadar reports on cybersecurity threats and incidents for informational and awareness purposes only. We do not engage in hacking activities, data exfiltration, or the hosting or distribution of stolen or leaked information. All content is based on publicly available sources.