Rosetta Daily · Jun 1, 2026

Auto-generated · 29 sources configured (WebSearch mode — feed allowlist still in effect) · 15 items selected
Window: past 24 hours (2026-05-31 → 2026-06-01)

In-the-Wild / Actively Exploited

• 🔴🔥 Microsoft Defender flaws actively exploited — CVE-2026-41091 & CVE-2026-45498 added to CISA KEV (deadline June 3)
  CISA added two actively-exploited Microsoft Defender vulnerabilities — CVE-2026-41091 (elevation of privilege) and CVE-2026-45498 (denial of service) — to the KEV catalog, with an FCEB remediation deadline of 2026-06-03. Fixed in Microsoft Defender Antimalware Platform 1.1.26040.8 / 4.18.26040.7 respectively; confirm the platform actually auto-updated rather than assuming it did.
  Malwarebytes · The Hacker News · CISA KEV

• 🔴🔥⚠️ PAN-OS GlobalProtect Authentication Bypass (CVE-2026-0257) — exploitation expanding, KEV deadline running
  Palo Alto Networks confirmed (advisory updated 5/29) that the GlobalProtect portal/gateway auth-bypass (CVSS 7.8) is under active exploitation to establish unauthorized VPN connections. CISA added it to KEV on 5/29; the federal remediation clock is ticking. Affects firewalls with auth-override cookies enabled under a specific certificate config. Patch fixed PAN-OS maintenance releases; disable auth-override cookies as interim mitigation.
  The Hacker News · Palo Alto advisory · CISA KEV

• 🔴🔥 FortiClient EMS Auth Bypass (CVE-2026-35616) — still being exploited to drop EKZ credential stealer
  Arctic Wolf continues to track exploitation of the pre-auth API access bypass (CVSS 9.1) in FortiClient Endpoint Management Server. Operators disguise the EKZ credential stealer as a Fortinet endpoint update and execute it via PowerShell, abusing trusted endpoint-management infrastructure to fan out across managed endpoints. Fixed in FortiClient EMS 7.4.7+; verify firmware-upgrade-reminder and Remote Access Profile configs were not tampered.
  The Hacker News

Critical Vulnerabilities & Advisories

• 🔴 SEPPMail Secure E-Mail Gateway — pre-auth RCE chain + mail-traffic access (CVE-2026-2743 CVSS 10.0, CVE-2026-44126/44128)
  InfoGuard Labs disclosed a cluster of critical flaws in SEPPMail Secure E-Mail Gateway: CVE-2026-2743 (CVSS 10.0, path traversal → arbitrary file write → RCE), CVE-2026-44126 (9.2, untrusted deserialization → unauth RCE), and CVE-2026-44128 (9.3, eval injection via the /api.app/template upldd param → unauth RCE). Exploitation lets an attacker read all mail traffic or pivot into the internal network; Censys shows several thousand public instances. Patch to 15.0.4 (earlier releases only partially fix the set).
  The Hacker News · InfoGuard Labs · CyberSecurityNews

• ⚠️ CIFSwitch — 19-year-old Linux kernel LPE to root (no CVE assigned yet, public PoC)
  Disclosed on oss-security (5/28) by Asim Manizada with a working PoC. cifs_spnego_key_type (in fs/smb/client/cifs_spnego.c) lacks a .vet_description hook, so the kernel never verifies that a cifs.spnego key request actually originated from the CIFS subsystem — an unprivileged user can impersonate a trusted kernel request and trigger privileged operations, reaching root. Latent since 2007. Exploitation needs a vulnerable kernel + compatible cifs-utils + unprivileged user-namespace creation. Upstream fix is commit 3da1fdf; fixed kernel versions vary by distro. Notably the bug was found via AI-assisted multihop reasoning over kernel semantic graphs — the third high-impact Linux LPE this cycle after CVE-2026-46333 (ptrace) and the Dirty Frag chain.
  CyberSecurityNews · securityonline · CloudLinux mitigation

Vendor Advisories

• Trend Micro Apex One — two critical RCE flaws patched (CVE-2026-34926 already in KEV)
  Trend Micro patched two critical Apex One RCE vulnerabilities on Windows. CVE-2026-34926 (directory traversal) was added to CISA KEV on 5/21 — patch now if not already done.
  Bleeping Computer

• Juniper Junos OS Evolved — unauth root RCE on PTX routers
  A critical Junos OS Evolved flaw on PTX Series routers allows unauthenticated remote code execution with root privileges. Patch per Juniper advisory; restrict management-plane exposure.
  Bleeping Computer

Web Security Research

• ⚠️ ChatGPhish — ChatGPT web-summary renderer turned into a phishing/prompt-injection surface
  Permiso Security's disclosure continues to circulate: the chatgpt.com response renderer trusts Markdown links/images from third-party pages ChatGPT summarizes, auto-fetching attacker-hosted images and surfacing links as clickable elements inside the trusted assistant UI — enabling indirect prompt injection and credential-phishing, and leaking IP/User-Agent/Referer. A clean illustration of "the summarizer is the attack surface."
  The Hacker News

• ⚠️ Marimo CVE-2026-39987 — pre-auth RCE chained with an LLM agent for post-exploitation
  Sysdig's writeup of an attacker exploiting a public Marimo notebook (pre-auth RCE, all versions ≤0.20.4) then driving an LLM agent to pull cloud credentials and exfiltrate a PostgreSQL database in under two minutes remains the standout "agentic post-exploitation" case of the week. Treat internet-reachable notebook/data-science tooling as crown-jewel-adjacent.
  The Hacker News

AI Security

• ⚠️ Microsoft MDASH — 100+ AI agents find 16 unknown Windows vulnerabilities (4 critical)
  Microsoft's MDASH harness orchestrates 100+ specialized AI agents; in Windows testing it surfaced 16 previously-unknown vulnerabilities, four critical enough to enable remote control. Continuation of the AI-finds-bugs-at-scale arc alongside Anthropic's Mythos — the disclosure-to-exploit window keeps compressing.
  SecurityWeek

• 🛡 Malicious browser extensions surge — QuickLens Chrome extension compromised to push malware / steal crypto
  A Chrome extension ("QuickLens — Search Screen with Google Lens") was pulled from the Web Store after being compromised to deliver malware and drain crypto from thousands of users. Part of a broader pattern of attackers leaning on extensions to steal cookies/tokens and log keystrokes. Audit installed extensions and enforce allowlisting.
  Bleeping Computer

Threat Intelligence

• Europol "Project Compass" — 30 arrests, 179 suspects tied to "The Com" cybercrime collective
  A year-long Europol-coordinated operation led to 30 arrests and tied 179 suspects to "The Com," an online collective known for targeting children and teenagers. Significant disruption of a sprawling social-engineering/extortion network.
  Bleeping Computer

• North Korea deploys new tooling to bridge air-gapped systems
  North Korean operators are using newly uncovered tools to move data between internet-connected and air-gapped systems via removable drives, plus covert surveillance capabilities — a reminder that USB/removable-media controls still matter for high-assurance environments.
  Bleeping Computer

• CISA details RESURGE implant from Ivanti Connect Secure zero-day attacks (CVE-2025-0282)
  CISA released new analysis of RESURGE, a malicious implant used in zero-day attacks against Ivanti Connect Secure. Useful IOCs/behaviors for anyone with Ivanti edge exposure in the hunt backlog.
  Bleeping Computer

Chinese-Language Community Picks

• FreeBuf / Anquanke / Xianzhi Communityand others continued to follow English-sphere stories this cycle: Microsoft Defender CVE-2026-41091/45498 exploited in the wild and added to KEV, PAN-OS CVE-2026-0257, the SEPPMail gateway RCE chain (CVE-2026-2743 and others), CIFSwitch Linux kernel privilege escalation, the FortiClient EMS EKZ stealer, plus high-severity patches for Trend Micro Apex One and Juniper Junos OS Evolved.
• Focus among domestic site operators and ops teams: once a secure mail gateway like SEPPMail is breached, the attacker can read all mail traffic, so upgrade immediately; FortiClient EMS is widely deployed for endpoint management in China, so verify that the firmware update path has not been tampered with.
• On the supply-chain front, the Chinese-language sphere continues to discuss Megalodon's GitHub CI/CD poisoning (5,561 repositories)and the ongoing TeamPCP/Shai-Hulud series.
  FreeBuf

Ransomware Today

• ~30–40 new DLS posts in last 24h.Most active groups: Qilin, Akira, DragonForce/Devman, The Gentlemen, Sinobi. Qilin sustains its lead with continued healthcare/legaltargeting; Akira active against manufacturing. Macro signal of the day: the ransomware payment rate has fallen to 28% — an all-time low— despite higher claimed-attack volume (Coveware-cited reporting), continuing the multi-quarter decline in victims who pay. Watchlist hits today: Qilin × healthcare, Akira × manufacturing.
• Full per-victim breakdown → intel/ransomware/daily/2026-06-01.html

Notable Breaches

• 600,000 affected by data breach at a healthcare firm(5/30); UFP Technologies(medical-device manufacturer) discloses an incident compromising IT systems and data; Olympique de Marseilleconfirms a cyberattack; ManoManonotifies customers of a breach via a compromised third-party provider; South Korea's National Tax Serviceaccidentally exposed a seized crypto wallet's recovery phrase, leading to ~$4.8M (6.4B won) theft.
  Bleeping Computer· SecurityWeek· CM-Alliance June 2026 tracker

---

AI Frontier

OpenAI

• No net-new model in the 5/31–6/1 window.Recent activity: Codex desktop-agent updates, reported confidential IPO filing prep (Goldman Sachs / Morgan Stanley). CNBC
• ChatGPhish— ChatGPT summary renderer abused for phishing + indirect prompt injection (see Web Security Research). The Hacker News

Anthropic

• Claude Opus 4.8 (5/28)— current flagship; stronger coding, agentic skills, and reasoning vs Opus 4.7, shipped in Claude Code with high-effort defaults. Releasebot
• Series H — $65B raised at $965B post-money (5/28)— led by Altimeter, Dragoneer, Greenoaks, Sequoia. Anthropic
• "Strengthening our model weight security" (5/28)— Anthropic published updated practices for protecting model weights against exfiltration. Anthropic
• Claude for Financial Services expansion (5/27)+ Code with Claude — Tokyo (Jun 5–6). Anthropic events
• Project Glasswing / Claude Mythos— 10,000+ high-severity flaws across widely-used software; first monthly update (5/22) still the reference point. Anthropic· The Hacker News

Google DeepMind / AI

• Exposed Google API keys in client-side code can authenticate to Gemini— embedded keys (e.g., Maps) could be reused to reach the Gemini assistant and access private data. Rotate and scope-restrict client-side keys. Bleeping Computer

🛡 = security-related

---

Failed Sources

• RansomLook API + RSS— still blocked by sandbox egress allowlist (WebFetch domain filtering). Ransomware section uses WebSearch aggregation of secondary trackers; per-victim accuracy not guaranteed.
• Chinese sources (FreeBuf, 安全客, 先知社区)— direct feed fetch not available in current sandbox; coverage based on WebSearch summaries.

---

Sources used: see intel/sources.yaml