SecurityMicrosoft Defender found malicious Chromium browser extensions masquerading as popular AI assistant add-ons that can harvest sensitive ChatGPT and DeepSeek prompts, responses, visited URLs, and internal browsing context, then quietly exfiltrate that data over routine-looking HTTPS traffic. The discovery matters because these extensions reportedly reached about 900,000 installs and appeared across more than 20,000 enterprise tenants, turning trusted browser marketplaces and everyday AI workflows into a significant data-loss risk for organizations.
3 min read · Mar 5, 2026
SecurityMicrosoft says the Tycoon2FA phishing-as-a-service kit has enabled attackers to bypass MFA at massive scale by using adversary-in-the-middle proxying to steal session cookies during real logins. That matters because organizations cannot rely on passwords and MFA alone to stop account takeovers—defenders must also revoke active sessions, harden authentication controls, and prepare for highly automated phishing campaigns that can reach hundreds of thousands of organizations.
3 min read · Mar 4, 2026
SecurityMicrosoft Defender Experts uncovered phishing campaigns that use fake meeting, invoice, and document lures to deliver malware installers disguised as trusted apps like Teams, Zoom, and Adobe Reader. What makes this notable is that the installers were signed with an EV certificate and used to deploy legitimate remote management tools such as ScreenConnect and Tactical RMM, giving attackers stealthy initial access that can blend into normal enterprise IT activity.
3 min read · Mar 3, 2026
SecurityMicrosoft says attackers are abusing a normal Microsoft Entra ID OAuth redirect behavior to turn trusted login links into phishing or malware delivery paths, often by forcing an OAuth error and sending victims to attacker-controlled redirect URIs. The campaign matters because it can bypass user suspicion and some security filters without stealing tokens, and it has been observed targeting government and public-sector organizations.
3 min read · Mar 2, 2026
SecurityMicrosoft says traditional threat modeling is no longer enough for generative and agentic AI apps because these systems are nondeterministic, easier to manipulate through prompt injection, and increasingly connected to tools, memory, and autonomous workflows. The guidance matters because it helps security teams anticipate AI-specific risks like tool misuse, privilege escalation, and silent data leakage before they turn into real-world exploits.
3 min read · Feb 26, 2026
SecurityMicrosoft Defender Experts uncovered a campaign using malicious Next.js repositories, often disguised as job take-home assignments, to compromise developers through normal VS Code and app startup workflows. The repos can automatically execute attacker-controlled JavaScript on folder open or when the project is run, potentially exposing source code, environment variables, and cloud credentials—making this a serious supply-chain and developer workstation security threat.
3 min read · Feb 25, 2026
SecurityMicrosoft has outlined 10 common Copilot Studio agent misconfigurations—such as oversharing, missing authentication, unsafe actions, and stale ownership—and paired them with Defender Advanced Hunting community queries to help security teams detect them. This matters because low-code AI agents are becoming a new control plane for identity, data access, and automation, meaning small setup mistakes can quietly expand an organization’s attack surface and enable abuse or data exfiltration unless proactively monitored and locked down.
3 min read · Feb 25, 2026
SecurityMicrosoft says its Defender XDR platform is evolving toward an “autonomous defense” model, combining unified cross-domain security signals, AI-driven investigation and response, and expert-led services to help SOCs move faster than attackers. The shift matters because many security teams are overwhelmed by fragmented tools and alert backlogs, and Microsoft argues this approach can reduce analyst toil, close visibility gaps, and improve early attack disruption.
3 min read · Feb 25, 2026
SecurityMicrosoft has released a new e-book, “Establishing proactive defense,” outlining a five-level maturity model for adopting Microsoft Security Exposure Management and moving from reactive vulnerability fixing to a continuous, risk-based approach. The guide matters because it gives IT and security teams a practical framework to unify visibility across hybrid environments, prioritize remediation based on business risk, and build a more measurable, effective security program.
3 min read · Feb 19, 2026
SecurityMicrosoft Defender is warning enterprises that self-hosted OpenClaw agents should be treated like untrusted code with persistent credentials, because they can ingest malicious text, install third-party skills, and act with the full trust of the host they run on. The guidance matters because it highlights a new blended risk model—where both software supply chain attacks and prompt injection can combine to compromise systems, identities, and sensitive data across multiple agents.
3 min read · Feb 19, 2026
SecurityMicrosoft’s new State of the SOC report, based on Omdia research, finds that security teams are being overwhelmed by fragmented tools, manual triage, and alert overload, with SOCs using an average of 10.9 consoles, nearly half of alerts being false positives, and 42% going uninvestigated. The findings matter because they show how tool sprawl and incomplete visibility are creating real operational and business risk, strengthening the case for unified SecOps platforms, automation, and AI-assisted workflows.
3 min read · Feb 17, 2026
SecurityMicrosoft Defender Security Research has identified 10 common Copilot Studio agent misconfigurations—such as overbroad sharing, missing authentication, risky HTTP actions, email-based exfiltration paths, and dormant connections—that can quietly expose organizations to serious security risks. Microsoft says these issues can now be proactively found through Advanced Hunting Community Queries in Defender, giving security teams a practical way to detect and fix dangerous agent setups before they are abused.
3 min read · Feb 12, 2026