Rosetta Daily · Jun 10, 2026

Generated automatically · 31 sources configured (WebSearch aggregation + RansomLook API direct) · 17 items selected
Window: past 24 hours (2026-06-09 → 2026-06-10). Wednesday — day after June Patch Tuesday. Headline themes: a critical Check Point VPN auth-bypass exploited as a 0-day to deploy ransomware (CISA KEV, 6/11 deadline), a fresh WordPress unauth-RCE under active attack, an AI-supply-chain twist (OpenClaw "skill" poisoning), and a notable LockBit5 resurgence burst on the leak sites.

In-the-Wild Exploitation / Actively Exploited

• 🔴🔥⚠️ Check Point VPN CVE-2026-50751 — CVSS 9.3 auth bypass, exploited as a 0-day since May 7 to deploy ransomware; CISA KEV, FCEB deadline 6/11
  A logic flaw in Remote Access / Mobile Access certificate validation lets an attacker establish a VPN connection without a valid password on gateways using the deprecated IKEv1 protocol with legacy clients and no required machine certificate. Check Point confirms active exploitation dating to May 7 with an early-June spike; at least one intrusion is linked (medium confidence) to a Qilin ransomware affiliate. CISA added it to KEV with a June 11 remediation deadline. A companion bug, CVE-2026-50752, enables MitM interference with site-to-site IKEv1 VPNs. Fix: apply Check Point's hotfix on an emergency basis; disable/replace IKEv1 legacy Remote Access where possible; require machine certificates.
  Check Point blog · Rapid7 · The Hacker News · BleepingComputer — CISA 3-day order

• 🔴⚠️ Everest Forms Pro (WordPress) CVE-2026-3300 — CVSS 9.8 unauthenticated RCE, actively exploited; full site takeover
  An unauthenticated eval() injection in the Everest Forms Pro plugin (~4,000 active installs) gives remote code execution and full site compromise. Affects all versions up to and including 1.9.12; active exploitation is ongoing. Fix: update immediately to the patched release; audit affected WordPress hosts for webshells and unexpected admin users.
  SecurityWeek / Hacker News aggregation · CISA KEV

• 🔥 SolarWinds Serv-U CVE-2026-28318 — CVSS 7.5 DoS, added to CISA KEV on evidence of active exploitation
  CISA added the SolarWinds Serv-U multi-protocol file-server flaw to KEV; a crafted request crashes the service (denial of service). Fix: patch Serv-U to the fixed build and restrict management-interface exposure.
  The Hacker News · CISA KEV

• 🔥⚠️ Microsoft Exchange CVE-2026-42897 — Critical spoofing flaw exploited; addressed via June Patch Tuesday after EEMS auto-mitigation (carry-over)
  The on-prem Exchange spoofing flaw that was being exploited with no patch is now folded into the June 9 Patch Tuesday cycle; until fully patched, the Exchange Emergency Mitigation Service (EEMS) remains the stopgap. Action: apply June Exchange updates; verify EEMS was never disabled.
  Zecurit — Patch Tuesday · Tenable

• 🔥⚠️ Microsoft Defender CVE-2026-41091 (EoP, public exploit) & CVE-2026-45498 (DoS, exploited) — carry-over KEV pair
  Defender elevation-of-privilege CVE-2026-41091 has public exploit code and the DoS CVE-2026-45498 is being exploited; both remain on the priority list heading out of June Patch Tuesday. Action: confirm Defender platform/engine updates have rolled to all endpoints.
  The Hacker News · CISA KEV

Critical Vulnerabilities & Advisories

• 🔴 June 2026 Patch Tuesday (6/9) landed — final window to complete Secure Boot certificate validation before the 6/26 deadline
  June's release is the last Patch Tuesday before the June 26, 2026 Secure Boot certificate hard deadline; 2026 cycles have averaged 100+ CVEs/month. Rogue researcher "Nightmare-Eclipse" (six Windows 0-days released since April: BlueHammer, RedSun, UnDefend, YellowKey, GreenPlasma, MiniPlasma) had promised "a big surprise" for this cycle. Action: prioritize Secure Boot cert rollout, Netlogon/Exchange/Defender exposures, and watch for post-PT 0-day drops.
  Zecurit — Patch Tuesday · Dark Reading — post-PT zero-days

• 🔴 n8n critical vulnerability — exposed instances at risk of full takeover
  SecurityWeek reports a critical flaw exposing internet-facing n8n automation instances to takeover. Action: restrict n8n to private networks / behind SSO, and update to the fixed version.
  SecurityWeek

Vendor Advisories

• Check Point — emergency hotfix for IKEv1 Remote Access VPN (CVE-2026-50751 + CVE-2026-50752); treat as emergency given confirmed 0-day exploitation and the 6/11 CISA deadline. Check Point
• Microsoft — June 2026 Patch Tuesdayrollup including the Exchange and Defender items above; pair with Secure Boot certificate validation before 6/26. Zecurit

Web Security Research

• PortSwigger Research — "Meet the HTTP Terminator" (Black Hat USA 2026) + Top 10 Web Hacking Techniques of 2025
  James Kettle's upcoming Black Hat talk asks whether an autonomous system can invent new attack techniques and hack live targets at scale, and reports new HTTP desynctriggers/gadgets that compromised banks, security products and government infrastructure. The 2025 Top 10 continues to headline malformed-chunk desync, browser-redirect-stalling chains, and SAML auth-bypass. Action:re-test desync and SAML exposure on internet-facing stacks.
  PortSwigger — Black Hat preview· Top 10 of 2025

AI Security

• ⚠️🛡 OpenClaw supply-chain poisoning — 341 malicious "skill" modules via ClawHub spread Atomic Stealer; payloads hidden in SKILL.md
  Chinese and Western reporting describes a supply-chain attack on the OpenClaw agent platform: 341 malicious skill modules distributed through ClawHub embedded payloads inside SKILL.md files to deliver Atomic Stealer, with a reported ~12% infection rate and links to ransomware crews. This is the agent-ecosystem analog of npm/pip poisoning — the "skill" is the new package. Action: treat third-party agent skills/MCP tools as untrusted code; review SKILL.md and tool manifests, pin/verify sources, and sandbox skill execution.
  FreeBuf weekly (CN-SEC mirror) · Microsoft — When prompts become shells

• ⚠️🛡 Prompt injection remains OWASP #1 — research synthesis: >85% attack success vs SOTA defenses under adaptive attacks; present in 73% of production AI deployments
  A review across 78 studies (2021–2026) finds adaptive prompt-injection attacks beat state-of-the-art defenses >85% of the time, with injection observed in 73% of production deployments; new arXiv work systematizes injection against agentic coding assistants (skills, tools, MCP). OpenAI has publicly conceded prompt injection in AI browsers "may never be fully patched." Action: assume injection is unsolved — enforce tool allowlists, least-privilege, human-in-the-loop on high-impact actions, and isolate model-output→interpreter paths.
  arXiv — agentic coding assistants · Vectra — prompt injection

Threat Intelligence

• ⚠️ Intel 471 — June APT surge across critical sectors: Silver Fox (Taiwan), PathWiper (Ukraine), Iranian APT33/34/39
  Intel 471's June update flags escalating espionage + disruption: Silver Fox spear-phishing Taiwanese government/tech with Gh0stCringe / HoldingHands RATs to steal IP; a new PathWiper destructive wiper hitting Ukrainian critical-infrastructure operators with time-based triggers and anti-recovery code; and Iranian groups (APT33/34/39) combining credential harvesting, wipers and ransomware against energy and infrastructure in North America, Europe and the Middle East. Action: hunt for the named RAT families; verify offline, immutable backups against wiper TTPs.
  Industrial Cyber — Intel 471

• ⚠️ Trend Micro — Pawn Storm deploys PRISMEX against government & critical infrastructure
  Trend Micro tracks Pawn Storm (Fancy Bear / APT28-nexus) delivering a malware family it calls PRISMEX in campaigns aimed at government and critical-infrastructure entities. Action: review detections for APT28 TTPs on perimeter mail and identity systems.
  Trend Micro · Dark Reading — Fancy Bear

Ransomware Today

• RansomLook API (direct): 32 fresh leak-site posts across 7 groups in the 24h window (2026-06-08 16:49 → 2026-06-09 14:50 UTC).Standout signal: a LockBit5 burst of 17 postsin a single ~17-second tranche (6/9 14:50 UTC) — consistent with the ongoing LockBit 5.0 resurgence(Q1 2026: #4, 163 victims, +106% QoQ). Qilinremained the steadiest poster (6); Akira, Stormous, RansomHouse, Termite, Novaalso posted. Watchlist group hits: lockbit5 (17), qilin (6), akira (2)= 25/32 posts. Note: today's RansomLook payload returned group + timestamp only (no victim names), so per-victim sector/geo detail is unavailable this edition.
• Full table + watchlist breakdown → intel/ransomware/daily/2026-06-10.html

---

AI Frontier

OpenAI

• GPT-5.5 Instant is the ChatGPT default— ~52.5% fewer hallucinated claims than 5.3 Instant on high-stakes (medicine/law/finance) prompts, at similar latency. TechCrunch
• Memory + personalization upgrade— ChatGPT can reference past conversations, files and Gmailfor personalized answers, and now shows memory sourcesacross models (provenance for generated answers); reviewable memory-summary page rolling out to US Plus/Pro. Releasebot — ChatGPT

Anthropic

• Claude Mythos released publicly as "Claude Fable" (6/9)— the model that, in restricted preview, autonomously discovered and chained 0-day exploits across major OSes/browsers, now shipped with enhanced misuse safeguards and priced ~2× current Opus tiers (institutional/security users). Gate News· TechCrunch — Mythos to critical infra
• Project Glasswing expansion— Mythos Preview to ~150 new orgs across power/water/healthcare/communications/hardware; Claude Securitydoes repo scans + patch suggestions. Releasebot — Anthropic
• Anthropic confidentially filed for IPO; financing round reported at a $965B valuation. CNBC

Google DeepMind / AI

• Gemini 3.5 Pro nearing June GA(2M-token context, "Deep Think"); 3.5 FlashGA since 5/19, default in the Gemini app & Search AI Mode. sources.news· blog.google — I/O 26
• Gemini Spark— a new general-purpose agent in the Gemini app that reasons across connected apps; beta for trusted testers / AI Ultra. CNBC

🛡 = security-relevant

---

Chinese Community Picks

• OpenClaw 供应链投毒— 341 malicious skill modulesvia ClawHubspreading Atomic Stealer(payloads in SKILL.md, ~12% infection rate, ransomware ties) is the standout local story this week, framing agent "skills" as the new supply-chain attack surface. FreeBuf 周报 (CN-SEC)
• Supply chain (供应链) stays the through-line— local follow-up on the codexui-androidnpm typosquat (exfiltrating persistent refresh_tokendisguised as Sentry traffic) and Gartner's six key 2026 cybersecurity trends. 安全内参 — FreeBuf
• Sysdig "first AI-driven cyberattack"(LLM agent, exposed marimo notebook CVE-2026-39987, ~22-min breach) continues to circulate as the reference case for AI-automated offense. FreeBuf

---

Failed / Limited Sources

• RansomLook /api/posts?days=1— reachable and used this edition, but the payload returned only group_name+ discovered(no post_title/victim/sector), so per-victim detail is omitted. rss.xmlreturned binary/non-text content via the fetch tool and was not parsed.
• ransomware.live /v2/recentvictims— blocked (URL not in fetch-provenance set); no cross-reference for victim names this run.
• Direct RSS/Atom feeds (CISA, The Hacker News, BleepingComputer, MSRC, PortSwigger, Unit 42/Mandiant, Chinese feeds)— direct fetch still constrained by the sandbox egress allowlist + provenance restriction; this edition was reconstructed via WebSearch aggregation over the same configured source set.

---

Source configuration: see intel/sources.yaml