Cyber Attacks & Data Breaches: August 2026
A month-end digest of ransomware, data breaches, zero-days, government incidents and a dedicated look at AI-driven attacks.
Read this first: this page is compiled from public reports and trackers. Counts differ between sources (leak-site claims vs confirmed attacks), attributions are often preliminary, and claimed breaches may be unverified. Always verify with the original advisory before acting.
🔒Ransomware & Extortion
August set a record for ransomware. Qilin (157 claims) and The Gentlemen (107) led, while Clop surged with 45 victims after exploiting PTC Windchill.
HostDZire (India)
A ransomware attack compromised multiple VMware ESXi nodes, causing wide outages and permanent customer data loss.
HostingCity of Lancaster, Ohio
Ransomware alerts appeared on multiple computers and the city took systems offline; about 90% of data was recovered by late August.
GovernmentFREYWILLE (Austria)
The jewelry maker disclosed an attack after detecting unauthorized activity four days earlier. Aurora claimed it and listed the firm.
ClaimedBelgian furniture retail chain
A brief disruption exposed customer names, addresses, phone numbers, emails and order histories.
RetailWindrose Health Network
Attackers exploited a flaw in a remote-management tool in early August, leading to a data breach.
HealthcareOVP Health
Newcomer group Storm claimed about 130GB of patient and employee data.
ClaimedCapgemini SE (claimed)
The Everest ransomware group was reported as claiming an attack on the consulting giant.
ClaimedBoston Scientific
The medical device maker reported on Aug 26 that it was hit by ransomware about a week earlier.
HealthcareShell and Philips
Both confirmed breaches tied to Clop’s PTC Windchill campaign, one of 40+ organizations targeted.
Clop campaign🗄️Data Breaches & Leaks
Third-party apps and single sign-on abuse drove the biggest breach of the month.
McKesson
ShinyHunters says it voice-phished staff, hijacked Okta SSO sessions and pulled data from Salesforce and Snowflake. It claims 284M records and demanded $55.2M; 6.4M emails leaked.
ShinyHuntersHCA Healthcare
A data breach affected about 55,000 people.
HealthcareMarathon Health
Reportedly affected 5.7 million individuals.
HealthcareEl Camino Health
A patient data breach involving four employee accounts.
HealthcareDreamHost
Attackers obtained FTP credentials through a file manager and may have accessed webmail session data.
HostingBOK Financial
Listed as a ShinyHunters victim in a cyber attack on the financial group.
Finance⚡Zero-Days & Exploits
Google counted a record 22 exploited zero-days in August, versus about 11 a month earlier in 2026 and 8 in 2025.
Microsoft Patch Tuesday
About 421 CVEs fixed, including the exploited afd.sys zero-day CVE-2026-68820, linked to Lazarus/DPRK campaigns. CVE-2026-62832 was public before a patch.
Zero-dayPTC Windchill / FlexPLM
Clop exploited unauthenticated RCE CVE-2026-12569 (patched June 17) against industrial firms, stealing CAD files and blueprints.
ClopSonicWall SMA 1000
Zero-days were reported exploited by ransomware affiliates.
Edge devicePaperCut NG/MF
Two chained zero-days (CVE-2026-81578, CVE-2026-82078; CVSS 8.8 and 9.4) allowed unauthenticated RCE. Emergency patches followed and CISA added them to KEV.
Zero-day🏛️Government & Public Sector
Berlin Senate network
Rhysida hit Berlin’s government network and stole sensitive data. Authorities refused to pay the ransom.
RansomwareUS ATF
The Bureau of Alcohol, Tobacco, Firearms and Explosives confirmed a cyberattack that officials called a major cyber event.
FederalFrench tax authority
A breach affected about 678,000 people and businesses.
Identity data🤖AI-Driven Cyber Attacks
August closed with AI agents running whole exploit lifecycles. Google says these tools still have limits, but speed and scale are rising fast.
PaperCut AI agent swarm
Starting Aug 31, an operator ran hundreds of AI agents (a DeepSeek model in a Codex harness) to build the exploit chain in about 4 hours. It compromised 440 servers at 395 organizations in 48 countries. GreyNoise published findings Sep 9.
Agent swarmGoogle: first end-to-end AI intrusion
GTIG described “exploit_agent”, which compared patched and unpatched code to find and exploit flaws without a human in the loop, and a PentAGI-based framework used from late August against Asian organizations.
AutonomousCrowdStrike Threat Hunting Report
AI-agent-triggered detection leads were growing 2.5 times faster than human-triggered ones.
Trend dataAI-automated influence campaign
Google found a platform using the Gemini API and other AI services to automate a campaign targeting public sentiment on regional political issues.
Influence opsLLMjacking
Mandiant saw stolen cloud credentials used to run AI models, with fraudulent charges above $100,000 in a single day.
LLMjacking- Speed: a working exploit chain was assembled in roughly four hours.
- Scale: one operator hit hundreds of servers across dozens of countries.
- Defense: patch internet-facing servers within hours of an advisory, and monitor cloud billing for AI misuse.
- Comparitech, Industrial Cyber, Ransomnews, rSecurity, BlackFog, SharkStriker and Konbriefing incident roundups
- TechRepublic, Veracode, CyberScoop, Help Net Security, Rapid7 and Canadian Cyber Centre (vulnerabilities)
- Cloud Security Alliance, Business Standard, CyberScoop and Tech Insider (AI-driven attacks)
Independent roundup. Not affiliated with any vendor named. Next edition: September 2026.
