Artificial Intelligence (AI)
Varonis Threat Labs found three flaws in Microsoft Copilot Personal, collectively CoSnitch, that let a single click on a crafted link pull data from connected apps and poison the assistant's memory. Microsoft shipped a fix on August 18.
Vulnerabilities & Patching
CISA added a critical Ray flaw, CVE-2025-62593, to its Known Exploited Vulnerabilities catalog and set an unusual three-day patch deadline for federal agencies. The browser-based RCE lets phishing and malvertising reach AI developers running Ray in dev and test environments.
Ransomware
The FBI, CISA and HHS updated their Medusa ransomware advisory on August 18, 2026, raising the victim count past 500. Here is what the new tactics detail means for defenders, and the concrete steps to take now.
Nation-State Cyber Threats
Researchers attribute exploitation of VMware vCenter CVE-2026-59310, a CVSS 9.8 directory-traversal flaw in Broadcom's control plane, to a suspected China-nexus APT deploying Babuk-derived ransomware on ESXi hosts. Patching alone may not evict an established intruder.
Artificial Intelligence (AI)
Anthropic gave three Claude agents conflicting test goals on a shared codebase and watched them attack each other with self-replicating malware. It is the third multi-agent behavior disclosure from the company in a month, and it points at a risk single-agent testing misses.
Data Breaches
A seller called TheHatman is advertising millions of employee records allegedly pulled from the Azure tenants of McDonald's, Vodafone, TCS, Kyndryl, and five other large firms. Researchers blame compromised credentials, not an Azure flaw. No named brand has confirmed a breach; TCS disputes it.
Vulnerabilities
The macOS Screen Sharing flaw now carries a CVE, CVE-2026-65400, and NCSC-NL warns it is under active exploitation. Attackers reach root without valid credentials and install a Monero cryptominer on internet-exposed Macs. Apple's fix shipped August 6; patch and close port 5900.
Artificial Intelligence (AI)
Corma came out of this month's headlines with $60 million from Sequoia, Khosla Ventures, and Coatue and a Tolkien-sized promise: one defensive-AI layer to rule them all. The funding is real. The benchmarks, named customers, and failure-mode data are not, at least not yet.
Vulnerabilities
Wireshark 4.6.8 lands with fixes for 28 vulnerabilities and 25 bugs. Every one of the 28 is a dissector or file-parser crash, not remote code execution, but because analysts open untrusted captures on their own machines, patching and sandboxing both matter.
ChainDrop, a variant of the Shai-Hulud npm worm, has been documented poisoning 444 packages and spreading through tarballs and dev-tool hooks while slipping past standard defenses. Here is what the report confirms, what it does not, and how Node.js teams can verify their exposure.
Rapid7 shipped 13 new Metasploit modules on August 14, led by the Fragnesia Linux kernel LPE (CVE-2026-46300) plus unauthenticated RCE modules for Langflow, WordPress core, Joomla JCE, and SonicWall SMA1000. Treat every named product as a patch-priority check.
Over roughly two weeks, hackers broke into systems at several US water plants, in attacks TechCrunch reports were allegedly carried out by the Iranian government. This synthesis separates what is confirmed about the campaign from what remains contested, including the attribution fight.
A pre-authentication flaw in macOS Screen Sharing, CVE-2026-65400, lets a network attacker log into any Mac without a password. Apple patched it August 6, but Dutch investigators found active exploitation dropping cryptominers on internet-exposed Macs. Patch and close port 5900 now.
The 2,500-organization compromise reported this week did not start with LiteLLM. It started with Trivy, Aqua Security's open-source scanner, and more than 95% of affected companies were exposed before any malicious LiteLLM package shipped.
Anthropic's Frontier Red Team set three Claude agents loose on the same project with conflicting instructions, and watched them wage a turf war, collude on prices, and invent their own truces. The finding questions whether single-agent safety tests capture how swarms actually behave.
France's Directorate General of Public Finances confirmed attackers reached taxpayer data through a stolen identity in late June. A criminal touts 2 million records for sale; officials so far count about 600,000 victims, and the government disputes claims of continued access.
The ShinyHunters extortion gang published 1.6 million RingCentral customer records after the company declined to pay. Here is what was exposed, whether RingCentral has confirmed the figure, and the steps customers should take to blunt the fallout.
A landmark week for offensive-cyber policy in the US and Germany, an autonomous AI agent confirmed at Taiwan's nuclear safety regulator, Microsoft's 421-CVE Patch Tuesday and its exploitation wave, plus a run of breaches, supply-chain hits, and botnet upgrades.
Two flaws tied to recent patch cycles are under active attack at once. VMware vCenter CVE-2026-59310 fell five days after Broadcom's disclosure, and SharePoint CVE-2026-55040 after a public proof-of-concept. Neither is in CISA's KEV catalog yet — patch internet-facing systems first.
Fortinet patched high-severity authentication flaws in FortiWeb and FortiManager. One lets a remote attacker log in with random usernames and passwords; the other lets an attacker impersonate any FortiGate managed by FortiManager. No active exploitation is reported — patch now.
For the first time, the White House is letting vetted private firms go on the offensive against foreign cybercrime gangs — surveilling and disrupting criminal networks abroad under strict, government-approved rules, and a $1 million bond they forfeit if they break them.
