Loading story…
Microsoft is introducing new security measures and features to address the growing use of AI agents within enterprise environments. The company announced Entra Agent ID, which extends Microsoft Entra's identity and access management (IAM) capabilities to AI agents, treating them with the same governance and controls as human users. This includes assigning unique identities to agents and applying conditional access, identity governance, and protection policies. The move is in response to the anticipated proliferation of both sanctioned and unsanctioned AI agents within organizations, aiming to mitigate risks associated with shadow IT and unauthorized agent activity. Additionally, Microsoft is rolling out a strengthened Content Security Policy for Entra ID sign-ins, restricting script execution to trusted Microsoft domains to prevent external script injection and cross-site scripting attacks during authentication processes.
Alongside these IAM enhancements, Microsoft is experimenting with new features such as the Agent Workspace for Copilot, which allows AI agents to access and manipulate files in isolated environments on Windows systems. While this feature is designed to improve productivity, it comes with significant security warnings. Microsoft documentation highlights risks such as cross-prompt injection, where malicious instructions embedded in files or applications could trick AI agents into performing unauthorized actions, including data exfiltration. To address these concerns, Microsoft emphasizes user supervision, least privilege access for agent accounts, and explicit user approval for sensitive actions. These developments reflect Microsoft's broader strategy to secure the expanding role of AI agents in enterprise IT while acknowledging the unique security challenges they introduce.
Mallory correlates global threat intelligence with your attack surface — know if you’re exposed before adversaries strike.
6 events from the most recent confirmed update back to the earliest known activity.
Microsoft announced it will enforce a stricter Content Security Policy for browser-based Entra ID sign-ins at login.microsoftonline.com in mid-to-late October 2026. The change will block external script injection and unsupported browser extensions or tools that inject scripts into sign-in pages.
Microsoft said Entra Agent ID, after its public preview, is expected to become commercially available in 2026. Pricing had not yet been finalized at the time of the announcement.
Microsoft introduced Entra Agent ID within Entra IAM to assign managed identities to AI agents and apply controls such as conditional access, governance, and protection. The feature is intended to help enterprises monitor and restrict agent behavior and is available in public preview.
In documentation accompanying Agent Workspace, Microsoft warned that agentic AI features create notable risks from cross-prompt and indirect prompt injection, including possible data exfiltration. Researchers also noted that access to folders like Downloads increases exposure to untrusted third-party content, while Copilot Actions can retain desktop screenshots for up to 30 days.
Microsoft introduced an experimental Windows feature for Windows Insiders that lets Copilot agents access local files through a separate Agent Workspace account. The feature is disabled by default and supports background tasks such as resizing photos, renaming files, and filling out forms.
PromptArmor recently published a proof of concept showing that hidden instructions in a spreadsheet could indirectly prompt-inject Anthropic's Claude for Excel. The demonstration highlighted how malicious file content can manipulate AI agents into unintended actions.
Vulnerabilities, threat actors, malware, products, organizations, and breaches Mallory has linked to this story.
4 references tracked. Mallory keeps watching after this page renders.
semperis.com
Open sourcezdnet.com
Open sourcebleepingcomputer.com
Open sourcescworld.com
Open sourceMap indicators from this story to your assets and identify affected systems in minutes.
Every observed campaign, victim, and pivot linked to actors named in this story.
Malware, exploits, and IOCs connected to the activity described here.
YARA, Sigma, and Snort rules deployed to your SIEM as soon as they’re published.
Get matching new stories delivered to your team as they break — not the next morning.
Ask questions about this story and take action on the answers.
Other stories Mallory is tracking that overlap on entities, sources, or topic labels.
Microsoft has announced a major upgrade to its Copilot AI integration in Windows 11, introducing a new feature called Copilot Actions. This feature enables AI agents to perform real tasks on users' local files and applications, moving beyond simple chatbot interactions to more advanced, agentic capabilities. Copilot Actions allows users to interact with their PC using natural language, voice commands, and visual cues, enabling the AI to open and close apps, update documents, organize files, book tickets, and send emails on behalf of the user. The feature is currently being tested by members of the Windows Insider Program and will eventually be rolled out to all Windows 11 PCs, regardless of whether they are classified as Copilot+ devices. Microsoft has emphasized the importance of security and privacy in the design of Copilot Actions, especially in light of previous controversies such as the Recall feature. To address potential risks, Copilot Actions requires explicit user permission before taking any action, and users retain the ability to control or revoke access at any time. Each AI agent operates within its own isolated Agent Workspace, ensuring that actions taken by one agent do not affect others or the broader system. Agents run under distinct standard Windows accounts with no administrative privileges, and their activities are governed by strict application and file system access rules. Microsoft has also implemented operational trust measures, such as digitally signed agents, and all agent activities are subject to the Microsoft Privacy Statement and Responsible AI Standard. The company is optimizing these workspaces to minimize any impact on system performance and has confirmed that there will be no changes to Windows 11 hardware requirements. The introduction of Copilot Actions represents a significant shift in how users interact with their PCs, transforming AI from a passive assistant to an active digital collaborator. However, this evolution also raises new questions about trust, as users must decide whether to allow AI agents access to sensitive files and applications. Microsoft is addressing these concerns by building in robust security guardrails and transparency around agent actions. The company is taking a cautious approach, learning from past experiences and ensuring that security controls are in place before the feature becomes widely available. As Copilot Actions moves from testing to general availability, Microsoft is expected to continue refining its security and privacy measures based on user feedback and ongoing risk assessments. The rollout of Copilot Actions is part of Microsoft's broader vision to deeply integrate AI into the Windows desktop environment, aiming to enhance productivity while maintaining user trust and system integrity. Security professionals and CISOs should closely monitor the deployment of Copilot Actions, assess the potential risks associated with agentic AI, and ensure that organizational policies are updated to address these new capabilities. The evolution of Copilot in Windows 11 underscores the growing importance of balancing innovation with security and privacy in the age of AI-driven computing.
Jun 29, 2026
Microsoft has introduced new agentic AI features in Windows 11, allowing background AI agents to perform tasks such as organizing files, scheduling meetings, and sending emails. These features, currently available as an optional toggle in Windows Insider test builds, are designed to enhance user productivity by enabling AI agents to act as digital collaborators. However, Microsoft has acknowledged that these agentic capabilities introduce novel security risks, particularly if attackers are able to manipulate the agents to execute malicious instructions or access sensitive user data. To mitigate these risks, AI agents are given separate user accounts and operate within a restricted workspace, but they may still require access to key user folders such as Documents, Downloads, and Desktop. Microsoft has issued explicit warnings that the agentic AI features should only be enabled by users who fully understand the security implications, and the setting is disabled by default. The company has published detailed support documentation outlining how these agents function, the permissions they require, and the potential for malware installation if the feature is misused. Security experts and Microsoft itself emphasize the importance of balancing productivity gains with the need to safeguard user data and system integrity as these advanced AI capabilities become more deeply integrated into the Windows operating system.
Jul 12, 2026
Major technology vendors and cybersecurity researchers are rapidly integrating artificial intelligence and automation into security operations, with Microsoft unveiling a comprehensive suite of AI-powered enhancements across its Defender ecosystem. These updates include proactive features such as Predictive Shielding for automatic attack disruption, a natural language Threat Hunting Agent, and expanded integration with third-party services like AWS and Okta. Microsoft is also addressing the growing challenge of non-human digital identities and agent sprawl, while expanding Security Copilot with dozens of new agents to automate tasks for security operations, identity, and IT teams. Meanwhile, the industry is seeing a surge in AI-driven detection engineering, with new and updated rules targeting advanced threats such as Windows defense evasion, credential access, phishing, and supply chain attacks. However, the adoption of generative AI models introduces new risks, as demonstrated by research into the Chinese DeepSeek-R1 model, which was found to generate insecure code—especially when prompted with politically sensitive topics. This raises concerns about the security implications of using foreign AI models, particularly those subject to state influence or censorship. Additionally, the threat landscape is evolving with the emergence of LLM-generated malware, adaptive AI-driven malware detection, and the use of AI in both offensive and defensive cyber operations. Security teams are urged to remain vigilant as AI technologies reshape both the tools available to defenders and the tactics employed by adversaries.
Jun 29, 2026