Rosetta Intel

Generated automatically · ~20 sources scanned · 21 items selected
Collection note: aggregated mainly via web search (most RSS/HTML feeds are not directly fetchable under sandbox/WebFetch provenance limits); RansomLook API and Bugcrowd CrowdStream JSON fetched directly. Failed sources at the end.

Critical Vulnerabilities

🔴⚠️ Check Point VPN Zero-Day Exploited in the Wild — Check Point Security Gateway / VPN, CVE-2026-50751, CVSS 9.3
Improper-authentication flaw lets an unauthenticated attacker establish a VPN session without valid credentials. Exploitation observed since ~May 7, 2026 with a spike in early June. Patch immediately and hunt for unauthorized VPN sessions.
Rapid7
🔴⚠️ Everest Forms Pro — Unauthenticated RCE — WordPress plugin, CVE-2026-3300
Actively exploited PHP eval() injection enables unauthenticated remote code execution. Update the plugin or pull it; WordPress plugin RCE chains are being weaponized fast.
securityonline
🔴 ServiceNow Unauthenticated API Exposure — ServiceNow hosted instances
A REST endpoint (/api/now/related_list_edit/create) shipped with requires_authentication=false, allowing unauthenticated querying of instance tables (tickets, employee records, incident reports). ServiceNow patched hosted instances on 2026-06-05; a confidential bug-bounty report had flagged it on 2026-04-22. ServiceNow attributes observed activity to researchers, not bad actors, and opened cases with affected customers.
BleepingComputer · The Hacker News · TechCrunch
⚠️ Gravity SMTP WordPress Plugin — Unauthenticated Info Disclosure — active on ~100,000 sites
Threat actors are exploiting an unauthenticated information-disclosure flaw to harvest sensitive data. Audit installs and update.
BleepingComputer
🔴 Google Chrome 149 — 33 fixes, 7 high-severity UAF (2026-06-18)
Use-after-free bugs can lead to remote code execution; stable channel updated to 149.0.7827.155/.156. Note: a separate V8 OOB read/write (CVE-2026-11645) is already on CISA KEV.
FreeBuf / gm7

In-the-Wild Exploitation (CISA KEV)

No new KEV additions in the past 24h. Most recent entries: CVE-2026-48907(Joomla Content Editor / Widget Factory JCE access-control, added 2026-06-16); the 2026-06-09 batch — CVE-2026-11645(Chromium V8 OOB R/W), CVE-2026-7473(Arista EOS), CVE-2026-20245(Cisco Catalyst SD-WAN Manager).
CISA 06-16· CISA 06-09
Ongoing this cycle (confirmed exploited): Check Point VPN CVE-2026-50751, Windows Netlogon RCE CVE-2026-41089, Oracle PeopleSoft CVE-2026-35273 (see prior briefs).

Vendor Advisories

Google Chrome— v149 ships 33 fixes incl. 7 high-severity use-after-free RCE vectors. gm7
Microsoft— June Patch Tuesday remains the largest ever (~200+ CVEs, 39 critical, 6 zero-days incl. exploited Netlogon). BleepingComputer
Check Point— emergency fix for actively exploited VPN zero-day CVE-2026-50751. Rapid7

Web Security Research

PortSwigger — "CSS: the bomb inside your inbox": multiple novel techniques to compromise email accounts by ripping apart trust boundaries with CSS/HTML, bypassing CSS sanitization, hardened CSP and HTML filters; demonstrated deanonymization of privacy-focused encrypted mail and end-to-end ATO on major providers. PortSwigger Research
Top 10 Web Hacking Techniques of 2025(19th edition) highlights malformed-chunk desync, browser-redirect stalling, and new SAML auth-bypass techniques. PortSwigger

AI Security

Mastra AI npm supply-chain attack → BlueNoroff/Sapphire Sleet— Microsoft attributes a supply-chain compromise of 140+ npm packages tied to the Mastra AI agent framework to North Korea's Sapphire Sleet (BlueNoroff). AI-tooling dependency chains are now a nation-state target. BleepingComputer
Prompt injection / jailbreak effectiveness stays high— systematic red-team eval: roleplay-based prompt injection ~89.6% attack success, logic-trap attacks ~81.4%, average successful jailbreak generated in <17 min for GPT-4; frontier models (GPT, Gemini, Claude, Grok) remain susceptible. Treat red teaming as a core dev practice. arXiv 2505.04806· getastra
Agent sandboxing converges— Anthropic (self-hosted sandboxes + private MCP) and Google (managed isolated Linux sandboxes) both ship agent execution boundaries this cycle, directly addressing agentic/MCP RCE surfaces. (See AI Frontier.)

Threat Intelligence

AryStinger router botnet— a previously undocumented malware botnet has compromised 4,000+ outdated routers, turning them into proxies for malicious traffic. BleepingComputer
Klue OAuth token theft → "Icarus" extortion— competitive-intelligence platform Klue confirms an incident where actors stole OAuth tokens used to connect to customers' Salesforce environments; the new "Icarus" extortion group claims it. Ties to the icarusgroup seen on RansomLook. BleepingComputer
'Prinz Eugen' ransomware— new operation prioritizes recently modified files for encryption and leaves no ransom note on the system. BleepingComputer
Texas Parks & Wildlife (TPWD) breach— vendor licensing-system breach exposed PII for 3M+ individuals. BleepingComputer
Cloud logging abuse— actors abuse AWS CloudTrail / Google Cloud Logging for defense evasion: stopping log collection, deleting log buckets, manipulating encryption keys to blind defenders. FreeBuf / gm7

Chinese Community

FreeBuf— Chrome high-severity flaws: v149 fixes 33 issues incl. 7 high-severity UAF enabling RCE (2026-06-18). gm7
FreeBuf— threat actors abuse cloud logging services (AWS CloudTrail / GCP Logging) to evade detection and degrade defender visibility (2026-06-18). gm7
FreeBuf— Anthropic's Project Glasswing reportedly helped early partners find 10,000+ high-severity vulns, now expanded to ~150 organizations. FreeBuf

Ransomware Today

RansomLook days=1 returned the same 6 posts (discovered 06-20/06-21) already covered on 06-22 — 0 net-new posts in the past 24h. The API appears stalled (replaying the prior batch). Watch icarus: its earlier "DEADLINE MONDAY" countdown persists and it is now linked to the Klue OAuth-token / Salesforce extortion (see Threat Intel).
Full table

Bug Bounty Today

5 recent disclosures (≤30 days) pulled from Bugcrowd CrowdStream JSON — all government VDP (NASA / GLOBE / FCC), no cash bounty but high P1 density and strong technique value. Highlight: a P3 reflected DOM XSS on NASA's VOTable Viewer chained via same-origin XHR to exfiltrate authenticated user PII (2026-06-17). HackerOne side not retrieved today (GraphQL/POST limits).
Full disclosures

AI Frontier

OpenAI

Frontier models + Codex now on AWS (06-02); GPT-5.2 retired in ChatGPT, sessions migrate to GPT-5.5 (06-12). Rosalind Biodefense (06-01); OpenAI Safety Bug Bounty (~$1M) still open for prompt injection / jailbreak / agentic risks. Releasebot

Anthropic

Claude Fable 5 released 06-09, then ordered offline 06-12by a US government export-control directive (also suspending Mythos 5) — export control first directly hits product availability. Seoul office (06-17). Claude Managed Agents now run in self-hosted sandboxes with private MCP. InfoQ

Google DeepMind / AI

Gemini 3.5 Flash GA; Managed Agents (Gemini API public preview) + Antigravity Agent run in isolated Google-hosted Linux sandboxes; Ultra cut $250→$200/mo. Google AI VRP pays up to $30K for AI-product flaws incl. prompt injection. Google Cloud

Failed / Not Directly Fetched Sources

Direct RSS/Atom/HTML (CISA KEV feed, NVD, PortSwigger, Project Zero, MSRC, The Hacker News, FreeBuf, etc.): not directly fetched (sandbox network restricted + WebFetch provenance gate); content aggregated via web search.
HackerOne Hacktivity: client-rendered + POST GraphQL — no per-item recent list retrievable via GET WebFetch today.

Sources used: see intel/sources.yaml

Rosetta Daily · Jun 23, 2026