CrowdStrike reports, in plain English
CrowdStrike publishes some of the most useful threat research in the industry. This page pulls out the headline findings from its 2026 Global Threat Report and Threat Hunting Report, explains its adversary names, and says what the figures mean for a Microsoft-centric organisation of a few hundred users.
2026 Global Threat Report
Published on 24 February 2026, covering activity in 2025. The headline is speed and quiet: attackers are moving faster and increasingly doing it without malware.
- Breakout time: the gap between first access and moving to another system. For eCrime it averaged 29 minutes in 2025.
- Malware-free: most detections involved valid credentials, trusted identity flows and approved SaaS integrations rather than malware.
- AI: CrowdStrike reports an 89% rise in attacks by AI-enabled adversaries compared with 2024, and says legitimate generative AI tools were abused at more than 90 organisations through malicious prompts.
- China-nexus: intrusions rose 38% across all sectors. CrowdStrike says 40% of the vulnerabilities exploited targeted edge devices, which often lack good monitoring.
- Cloud: cloud-conscious intrusions rose 37%, and valid account abuse accounted for 35% of cloud incidents.
- Social engineering: incidents using fake CAPTCHA lures rose 563%. That is the technique behind ClickFix.
- Adversary tracking: 24 new adversaries were named in 2025, taking the total tracked to more than 281.
2026 Threat Hunting Report
Published on 3 August 2026, covering the first half of 2026. The theme is a shrinking exploitation window and identity-led attacks.
- Exploitation window: CrowdStrike expects timelines to keep shrinking, partly because AI may speed up vulnerability discovery and exploit development.
- Vishing and single sign-on: voice phishing was used to steal data from SaaS applications and compromise single sign-on accounts. In one case an account takeover led to data theft in under five minutes.
- Device code phishing: monthly attempts rose 15 times over the past six months.
- AI agents: detection leads triggered by AI agents were 2.5 times more numerous than manually driven activity, which makes malicious and expected AI behaviour harder to separate.
- Supply chain: one group compromised more than 300 software dependencies in a single day, harvested credentials and moved into cloud environments.
- Scope: CrowdStrike now tracks more than 290 named adversaries.
My take for a Microsoft-centric estate
The two reports point the same way. For an organisation of about 500 users on Entra ID, Intune and Microsoft 365, I read the practical priorities like this:
- Plan for minutes, not hours: a 29-minute average breakout means detection and containment need to be continuous. Round-the-clock monitoring and pre-agreed isolation actions matter more than a good report afterwards.
- Treat identity as the perimeter: with 82% malware-free activity, the controls that count are phishing-resistant MFA, Conditional Access, short token lifetimes and sign-in monitoring. See adversary-in-the-middle token theft.
- Block what you do not use: if nobody legitimately uses device code sign-in, a Conditional Access policy that blocks it removes a technique that is growing quickly.
- Harden the help desk: vishing succeeds where identity checks are weak. Callbacks to known numbers and strict verification before any reset or MFA change are cheap defences.
- Know your edge: inventory VPNs, firewalls and other internet-facing devices, make sure they are logged, and patch internet-facing critical flaws in hours where you can, not weeks.
- Watch dependencies: if anyone builds software, pin and review packages, because registry attacks are mostly npm.
For wider context see the threat landscape, emerging threats and intel.
Reading the adversary names
CrowdStrike names the groups it tracks with a two-part label: a descriptive word followed by a suffix that says what kind of actor it is or where it is linked to. Other vendors use their own names for the same activity. Microsoft, for example, uses weather names, as in the Midnight Blizzard incident file.
| Suffix | What it denotes | Basis |
|---|---|---|
| Spider | Cybercriminals motivated by money | CrowdStrike naming article |
| Bear | Nation-state actors linked to Russia | CrowdStrike naming article |
| Panda | Nation-state actors linked to China | CrowdStrike naming article |
| Jackal | Hacktivists seeking political disruption | CrowdStrike naming article |
| Kitten | Groups linked to Iran | Grouping in CrowdStrike's adversary explorer |
| Chollima | Groups linked to North Korea | Used for "North Korea-nexus" groups in the Global Threat Report |
| Bison | Groups linked to Belarus | Used for a "Belarus-nexus" group in the Threat Hunting Report |
The naming article that defines the first four dates from 2022. CrowdStrike's adversary explorer lists the tracked groups by country or category, with more detailed profiles for customers. A "nexus" label means a link to a country, not proof of who is behind a specific incident.
Sources
- CrowdStrike: 2026 Global Threat Report: The Evasive Adversary Wields AI, 24 February 2026
- CrowdStrike: 2026 Threat Hunting Report, 3 August 2026
- CrowdStrike: Naming adversaries and why it matters to security teams, 31 May 2022
- CrowdStrike: Adversary explorer, undated