Rosetta Intel

Generated automatically · 27 sources scanned · 24 items selected

Critical Vulnerabilities

🔴🔥 SharePoint Server RCE (CVE-2026-45659) — Microsoft SharePoint, CVSS 8.8
Deserialization-of-untrusted-data RCE, added to CISA KEV after active exploitation. FCEB remediation was due July 4, 2026.
The Hacker News
🔴🔥 Adobe ColdFusion Path Traversal (CVE-2026-48282) — Adobe ColdFusion, CVSS 10.0
Path traversal enabling arbitrary code execution; part of CISA's July 7 KEV batch (Adobe/Joomla/Langflow), FCEB fix due July 10.
The Hacker News
🔴🔥 Joomla SP Page Builder Upload Flaws (CVE-2026-56290 / CVE-2026-48908) — JoomShaper / Joomlack Page Builder, CVSS 10.0
Improper access control + unrestricted dangerous file upload → unauthenticated RCE. Both added to KEV July 7.
The Hacker News
⚠️🔥 Langflow Authorization Bypass (CVE-2026-55255) — Langflow (AI workflow builder), CVSS 6.1
Authenticated attacker can execute another user's flows; added to KEV July 7 on exploitation evidence.
CISA

In-the-Wild Exploitation (CISA KEV)

CISA added twomore KEV entries on July 10, 2026(details still propagating; verify against the KEV catalog).
CISA alert
Recent KEV context: PTC Windchill/FlexPLM (CVE-2026-12569) and Cisco Unified CM SSRF (CVE-2026-20230) added late June.

Vendor Advisories

Cisco (July 1 batch)— Catalyst Center Arbitrary File Read (CVE-2026-20191, CVSS 7.5) and ClamAV flaws affecting multiple Cisco products (High).
Cisco advisory notice
Fortinet / Ivanti (recent)— Critical OS command injection and auth-bypass fixes remain worth verifying in-fleet (e.g., Ivanti CVE-2026-10520 CVSS 10, CVE-2026-10523 CVSS 9.9).
SecurityWeek

Web Security Research

PortSwigger — Black Hat / DEF CON USA 2026 previews— a "hacking hat-trick" of upcoming publications was teased; watch the Research feed for release. Reference back-catalog: Static Path Deceptionand the Top 10 Web Hacking Techniques of 2025.
PortSwigger Research

AI Security

Indirect prompt injection still unsolved in agent contexts— Axis Intelligence Q2 2026 testing (312 attacks) reports agent-context attack success rates flat-to-worse across all six frontier models despite better single-turn jailbreak resistance; ~73% of production AI deployments show prompt-injection exposure.
Axis Intelligence
EchoLeak (M365 Copilot)and CVE-2025-53773 (GitHub Copilot RCE, CVSS 9.6)remain the canonical zero-click / hidden-injection references; OpenAI's "Lockdown Mode" (Feb 2026) concedes AI-browser injection "may never be fully patched."
Cycode

Threat Intelligence

Mandiant M-Trends 2026— median time from initial access to hand-off to a secondary actor collapsed to 22 seconds(from 8+ hours in 2022); "prior compromise" is now the top initial vector in ransomware (30%), while email phishing fell to ~6% of intrusions.
Google Cloud Blog
APT41 (DUSTTRAP)sustained campaign across shipping/logistics, media, tech and automotive (IT, ES, TW, TH, TR, UK); Iranian "Screening Serpens"expanding to US/Israel/UAE targets.
Unit 42

Chinese-Language Community Picks

FreeBuf — BitUnlocker: a BitLocker downgrade attack— exploits the gap between patch rollout and certificate revocation to physically crack the encrypted volume of a patched Windows 11 device within five minutes; derived from one of four 0-days found by Microsoft's STORM team (CVE-2025-48804).
FreeBuf
Pwn2Own Berlin 2026 registration full— capacity reached for the first time in 19 years, with rejected researchers launching "retaliatory disclosure"; the xchglabs team had prepared 86 vulnerabilities targeting NVIDIA, Docker, Linux KVM and PyTorch.
FreeBuf

Ransomware Today

16 new leak-site posts (07-10 → 07-11) across 7 groups. Qilin dominated with 8 victims (real estate, hotels, construction, BPO, finance; LatAm + APAC); Interlock hit Borger ISD (US K-12). Watchlist hits: 9 (qilin ×8, akira ×1). Off-board: Deutsche Bank breach claimed by "unsafe" (bank denies corporate-network compromise); Accenture confirmed an intrusion after actor "888" claimed 35GB of source code + cloud keys.
Full victim table

Bug Bounty

The Bug Bounty track now updates daily on its own (deep dives + themed recent disclosures).
Open the Bug Bounty daily track

AI Frontier

OpenAI

GPT-5.6expected imminently (benchmarked for speed/reliability/accuracy); OpenAI also announced a custom "Jalapeño" inference chip. Security-relevant: Lockdown Modefor ChatGPT (Feb 2026).

Anthropic

Claude Sonnet 5shipped — near-Opus-4.8 quality at intro pricing ($2/$10 per M through Aug 31), with autonomous browser/terminal tool use. (agentic tool use raises injection/abuse surface)

Google DeepMind / AI

Gemini 3.5 Pro delayed to July 17— full architectural rebuild (2M-token context, "Deep Think" reasoning layer, autonomous workflows). Also NanoBanana 2 Liteimage model (<4s, from $0.034/1k).

🛡 = security-relevant

Failed Sources (if any)

None hard-failed. Feeds were assembled via WebSearch per category (direct RSS/HTML fetch is provenance-gated in this environment) + RansomLook API for ransomware. Feed-level timestamps are approximate.
CISA July 10 KEV entry-level CVE detail not yet fully indexed — flagged for verification against the KEV catalog.

Sources used: see intel/sources.yaml

Rosetta Daily · Jul 12, 2026