Microsoft Digital Defense Report 2026
In an era of interconnected risk, attackers are increasingly exploiting the trusted identities, systems, relationships, and services organizations rely on most. The 2026 Microsoft Digital Defense Report reveals a fundamental shift in cybersecurity with the interconnected digital ecosystem and the rise of AI. This report helps organizations decode what is needed to secure AI and build resilience across the interconnected systems they depend on.
Share
Our unique vantage point
Microsoft’s global presence—spanning billions of users, millions of organizations, and a vast network of partners—provides us with an unparalleled perspective on the cybersecurity threat landscape.
165+ trillion
Security signals processed daily
15,000+
Partners in our security ecosystem, making it one of the largest in the world
4.7 million
Net new malware file blocks every day
35,000
Full-time equivalent security engineers employed worldwide
31 million
Identity risk detections analyzed in an average day
5.2 billion
Emails screened daily on average to protect users from malware and phishing
Top 10 takeaways from this report
AI is changing the physics of cybersecurity
AI is compressing attack timelines, lowering the cost of sophisticated capabilities, and enabling attackers to operate with greater speed, scale, and autonomy. But the same capabilities can strengthen defense by accelerating discovery, analysis, prioritization, and response. Security organizations must prepare for an environment in which both attackers and defenders increasingly operate at machine speed.
Attackers are already targeting AI as another attack surface
Threat actors can exploit AI to do things like execute malicious commands, steal AI serving capacity, and exfiltrate data. As AI becomes ubiquitous in the workplace and at home, it has become both a tool and a target for attackers.
People remain a heavily exploited initial-access path
Threat actors continue to target user behavior through user execution, ClickFix-style social engineering, phishing, and impersonation. These techniques remain effective even as AI changes their speed, scale, and personalization.
Identity is the primary control plane for defense
Attackers consistently take advantage of unnecessary privileges, persistent access, and weak identity governance. This is why phishing-resistant multifactor authentication (MFA) and passkeys, disciplined identity hygiene, tiered administration, and strong privileged-access enforcement remain the best safeguards against cyber attacks. That control plane now spans both human and non-human identities, including applications and agents.
Signal correlation through shared threat intelligence is key to defense at multiple levels
The degree to which a defender’s systems share and correlate threat intelligence is one of the highest-leverage strategic variables under the defender’s control. Signal cross-correlation can reveal attack patterns that remain invisible when endpoint, identity, cloud, application, email, and network telemetry are analyzed in isolation.
Emerging threats facilitate stealthier attacks, operating at a larger and faster scale
Over the next year, Microsoft sees the most significant threats coming from:
- Open-source supply chain compromise
- Attacks against edge devices
- AI as a force multiplier for malicious activity
These threats are emerging as AI-generated synthetic content erodes confidence in digital communications. As authentic and synthetic content become harder to distinguish, individuals can become simultaneously more vulnerable to manipulation and less confident in legitimate sources. This shifts the security model from implicit trust toward explicit verification.
Protecting data is even more important as AI expands access at scale
Trusted AI depends on trusted, well-governed data. Data protection has always been a key priority, and now that AI systems and agents can reach and act on sensitive information at unprecedented scale, it’s more crucial than ever. Organizations should reduce oversharing, apply sensitivity-aware access and least-privilege controls, and govern how data is discovered, used, and shared across AI, cloud, and application environments.
The future of defense involves rethinking how we protect our assets from reactive incident response to proactive, continuous discipline
Defenders must shift from tool-centric vulnerability management to threat exposure management. At the core of this is:
- Asset visibility
- Continuous vulnerability awareness
- Intelligence-driven threat detection
- AI-accelerated insight
This means reporting metrics should shift from number of patches deployed to exposure reduced, detection coverage increased, and time-to-mitigate compressed.
National and international policy decisions will shape the future of cybersecurity
In an increasingly contested landscape, cybersecurity demands clearer rules, credible accountability, and sustained cooperation. Strengthening protections for critical infrastructure, advancing operational accountability, and preserving an open, secure, and resilient digital environment must be a priority for governments on both a domestic and international level.
Resilience must be a core outcome of cybersecurity planning
Retain the existing focus on reducing exposure, detecting threats faster, adapting to changing risks, and maintaining operations. This preserves resilience as the outcome without competing with identity, secure foundations, data, and the three-part operating model for ‘primary’ status.
Key figures from the report
63%
Of intrusions involved data theft
5.3 hours
Average time before exposed cloud workloads were attacked
93%
Of voice phishing attacks kept the victim on the line long enough to begin social engineering
46 million+
Business email impersonation attacks detected over the past 12 months
89-95%
Of email phishing attachments led to a credential theft effort
Cyber threats: Worldwide customer impact in 2026
Countries where customers are most frequently impacted by cyber threats.
Source: Microsoft Threat Intelligence
This map pulls from data from July 1, 2025 – June 30, 2026 on how frequently customers are targeted by malicious activity in each country. The most impacted countries are compared to other countries in their region, both as a percentage of regional activity and a rank of regional activity.
MDDR 2026 Report
AI
AI
As AI becomes foundational to economies, governments, and critical services, organizations must secure the AI ecosystem through governance, accountability, lifecycle-based security, and resilient architecture.
The AI threat landscape
The expanded threat landscape includes both AI systems as targets and AI-enabled attacks on traditional systems. Attackers may target AI-specific components—including inputs, model data and weights, serving infrastructure, user data, and generated outputs—to manipulate, steal from, or disrupt AI systems. At the same time, AI can increase the speed, scale, and sophistication of attacks against more traditional targets, including computer systems, people, sensitive information, and physical capabilities.
Source: This figure maps the expanded target landscape, which roughly divides into targets associated with AI systems and targets of threat actors that use AI for enablement.
88%
88% of enterprises are already experimenting with AI agents, with 82% of leaders planning broader rollouts within 12 to 18 months.
1.3 billion
Industry projections point to roughly 1.3 billion AI agents in production by 2028.
Malicious AI assistant extensions harvesting LLM data
In December 2025, Microsoft found a malicious browser extension (600,000+ installs) harvesting ChatGPT and DeepSeek conversation history. This activity affected almost 10,000 organizations before it was mitigated. LLM conversation history now routinely holds source code, architecture, customer data, and leaked credentials.
The five risk classes defining the 2026 agentic attack surface
The chart identifies five key attack surfaces that could be used to exploit AI agents, pairing common tactics with the defensive controls needed to address each risk. Together, they show that securing agentic AI requires protections not only around the model itself but also around the identities, data, tools, and systems that agents can access and act on.
| Risk area | What attackers do | What defenders need |
|---|---|---|
| Prompt and intent manipulation | Inject instructions through prompts, content, files, or memory to redirect the agent | Prompt injection detection, payload inspection, intent validation, output review |
| Sensitive data exposure | Drive the agent to read or return content beyond its sanctioned scope | Sensitivity-aware retrieval, response inspection, data loss prevention (DLP) on inputs and outputs, memory scoping |
| Identity and privilege compromise | Impersonate a sub-agent, reuse stolen credentials, or chain privileges across agents | Verifiable agent identity, mutual authentication, scoped credentials, least privilege |
| Excessive agency | Coax the agent into chaining tools or actions beyond sanctioned behavior | Tool allow-listing, runtime gating, action policy, anomaly detection |
| Operational integrity | Tamper with configuration, system prompts, memory, training data, supply chain, or logs | Immutable logs, signed configuration, change governance, supply chain attestation |
Govern agent identity before agents outnumber people
AI is expanding what attackers can do and what defenders must protect at the same time. Policymakers should treat AI systems as infrastructure that requires security across the full lifecycle, enabling governments and organizations to adopt AI confidently while maintaining control over sensitive systems and data.
MDDR 2026 Report
ThE THreat landscape
The threat landscape
Most intrusions still start with a person or a credential rather than an exploit. Attackers are increasingly relying on human behavior and identity-based access, making intrusions harder to detect. This underscores the importance of foundational security and defense at machine speed.
Human action and trusted access still account for the majority of observed initial access
People and identities are primary entry points for attackers, reinforcing the need for layered identity, endpoint, and browser protections. Below are the most common initial access techniques identified in the 2026 report by share of observed activity.
52.2%
Of valid account intrusions involved follow-on credential theft. One compromised identity can fuel the compromise of many more identities.
46+ million
Business contact impersonation (BCI) attacks detected over the past 12 months.
25.5%
Of observed cyber threat activity impacted customers in the United States from January 2025 – June 2026.
Top 10 global sectors impacted by threat actors in 2026
Attackers prioritize targets (e.g., government, IT, and research and academia) with high-value intelligence, extensive personally identifiable information, often under-monitored infrastructure, and a generally low tolerance for downtime.
Nation state and geopolitical threats
Nation-state cyber activity is increasingly focused on gaining and maintaining trusted access to critical systems, identities, and digital ecosystems. Across China, Iran, North Korea, and Russia, state sponsored cyber operations continue to evolve toward more persistent, scalable, and strategically aligned access models, with a shared emphasis on long term positioning inside high-value environments.
Strengthen identity verification and reduce reliance on implicit trust
Policymakers should push organizations to treat identity as a core security control, limit standing and privileged access, and require phishing-resistant authentication. They should also encourage faster remediation of internet-facing vulnerabilities and stronger cross-system visibility so attackers cannot use trusted credentials or fragmented security gaps to move undetected across an environment.
MDDR 2026 Report
Cybercrime
Cybercrime
Cybercrime operates through interconnected criminal ecosystems that provide infrastructure, access, services, and expertise at scale.
As these markets mature, and with the rise of AI, cybercrime has become increasingly resilient, adaptable, and transnational, requiring coordinated action across governments, industry, and law enforcement.
Scale, speed, and automation are changing the game
While the tactics behind identity compromise remain largely unchanged, attackers are operating at greater scale and speed through automation and orchestration. Below is how the identity threat landscape has evolved, according to observations from the Microsoft Detection and Response Team (DART).
| Unchanged | Changed |
|---|---|
| Identity compromise | Scale of access |
| Credential reuse | Speed of execution |
| Privilege escalation | Increase of use of non-human identities |
| Access misuse | Automation and orchestration |
145+ million
QR-code phishing attacks detected between July 2025 and June 2026 by Microsoft Defender for Office 365.
89-95%
Of email phishing attachments led to a credential theft effort.
15.8%
Increase year-on-year in ransom detonations against enterprises.
Ransomware: Impacted industries
Ransomware affects nearly every industry, with critical manufacturing the most heavily targeted sector by number of attacks. Below is the year-on-year change in ransomware detections across industries, according to Microsoft Threat Intelligence data.
Disrupting Tycoon2FA
In March 2026, Microsoft collaborated with international law enforcement and industry partners to target the infrastructure supporting the phishing service. The team disrupted the servers where phishing pages were rendered and credentials were intercepted. By June, Tycoon2FA activity had fallen 95% from its November 2025 peak, while its share of CAPTCHA-gated phishing dropped from 76% to 12%.
Disrupt the ransomware economy through coordinated policy and enforcement
Governments should target the broader ransomware ecosystem by discouraging ransom payments, increasing accountability for safe-haven jurisdictions, strengthening cross-border enforcement, and imposing visible consequences through arrests, asset seizures, and infrastructure takedowns. Sustained public-private collaboration and intelligence sharing are critical to making these tools effective.
MDDR 2026 Report
Resilience
Resilience
Society depends on interconnected systems—from power and water to hospitals, schools, software supply chains, and subsea cables—so failures can quickly spread across sectors. The answer is not isolation, but building resilience across shared dependencies before a crisis happens.
78%
Of observed attack techniques against critical infrastructure used cloud identity abuse.
60-70%
Organizations leveraging Microsoft Security Copilot report summarizing threats 60 to 70% faster, sharply cutting the mean time to resolve critical alerts.
Attacks against state and local government, education (SLED) and the critical infrastructure sector
Research and academia is the most targeted SLED sector by a significant margin, accounting for 38% of observed attacks. Below is the breakdown of attacks across sectors, according to Microsoft Threat Intelligence data from November 2025 to April 2026.
Advancing Regional Cybersecurity (ARC) initiative
Microsoft’s Advancing Regional Cybersecurity (ARC) initiative uses scenario-based exercises and public-private collaboration to identify gaps in government coordination before an incident occurs. In Mexico, the work brought technical, civilian, and regulatory stakeholders together to test whole-of-government crisis response in preparation for the FIFA World Cup.
Build for continuity, not just prevention
Governments should treat resilience as a shared capability, investing in cross-agency coordination, common monitoring and response capacity, infrastructure redundancy, and regular exercises so essential services can continue when disruption occurs.
Additional report topics
Take a deeper dive into our 2026 Microsoft Digital Defense Report. Navigate to specific sections of the report below.
Introduction
AI
The Threat Landscape
Cybercrime
Resilience
Beyond our report
Follow us