⚠️ OpenAI says Astra found and used two zero-days in an evaluation exploit chain.

Google’s Gemini 3.8 Flash Cyber and Anthropic’s Claude 5.1 models also arrived with tiered access and safeguards.

What each company disclosed: https://thehackernews.com/2026/09/google-anthropic-and-openai-unveil.html

⚠️ Attackers are exploiting LiteLLM, Artifactory, and Switchvox flaws.

The attacks deploy crypto miners and reverse shells, mint admin tokens, and harvest API keys. CISA added seven exploited vulnerabilities to its KEV catalog.

Inside the attack chains: https://thehackernews.com/2026/09/cisa-adds-seven-exploited-flaws-as.html

‼️ A public PoC shows how CrowdStrike Falcon’s macro remediation can be abused for privilege escalation.

FalconFlank works on fully updated Windows 11 25H2 and Windows Server 2025. CrowdStrike has not yet commented.

Read: https://thehackernews.com/2026/09/researcher-releases-falconflank-poc.html

🚨 Pegasus infected a Serbian student movement member’s iPhone through a zero-click iMessage exploit.

At least 14 people in Serbia have faced advanced spyware targeting in 2026.

Read: https://thehackernews.com/2026/09/pegasus-zero-click-spyware-exploit.html

Shai-Hulud now checks 469 locations for credentials.

Earlier variants checked 189. The worm now scans developer environments, CI/CD tooling, cloud configs, and AI tool settings. Stolen publishing tokens can provide a path into trusted package distribution.

Inside the expanded credential hunt: https://thehackernews.com/2026/09/shai-huluds-reach-just-grew-to-469.html

⚠️ Attackers are abusing signed Node.js binaries to run malware.

Observed activity targeted government departments, tech companies, and hotels. In one intrusion, attackers used ClickFix for initial access and EtherHiding to retrieve commands.

How the attacks work: https://thehackernews.com/2026/09/attackers-turn-trusted-nodejs-runtime.html

🚨 US SOCs face a persistent risk: attackers move faster than detection rules.

Close the blind spots with live threat intel from 16K companies worldwide.

Get fresh IOCs in your SIEM within 24h after threats appear.

Integrate actionable TI: https://thn.news/threat-feeds

Organizations are increasingly relying on AI-generated and third-party code across the software supply chain. But once that code is compiled into a stripped binary, much of the original source context disappears, leaving security teams with limited visibility into what’s actually running and unable to detect malicious code entering production systems.

RevEng.AI is bridging that gap with Mega Bite, a suite of specialized AI models purpose-built for binary analysis. It includes two new proprietary models, WilBERT and Ventris, designed to analyze binary software without requiring access to source code while uncovering context that traditional decompilers and general-purpose LLMs can miss.

Check out the blog and watch this video to see how AI-powered binary analysis can give security teams deeper visibility into the software they run, with the speed and reliability to match real-world security operations.

https://thn.news/binary-analysis

‼️ ALERT - Thomson Reuters’ C-Track court software breach may have exposed SSNs and sealed court data.

An unauthorized party obtained files tied to courts across the U.S. and Ontario. Some courts say backup data was involved; Ohio says its production platform was accessed.

Details: https://thehackernews.com/2026/09/thomson-reuters-court-software-breach.html

⚠️ Criminals can buy access to Windows hosts compromised by BraZetsu.

The Python malware uses AI to prioritize high-value systems, collects browser and financial data, and lets buyers deploy their own payloads through purchased access.

Read: https://thehackernews.com/2026/09/brazetsu-malware-turns-compromised.html

🚨 An RMM phishing campaign first tied to Canada spans 46 countries.

ANYRUN linked 601 cases to the operation, with 45% of observed activity associated with the U.S. Fake tax, shipping, and invoice lures trick targets into installing legitimate RMM software.

How the chain works: https://thehackernews.com/2026/09/us-becomes-top-target-in-rmm-phishing.html

🚨 A critical Cisco Nexus 9000 flaw lets unauthenticated remote attackers run code as root.

CVE-2026-20212 leaves TCP ports 43210 and 43211 reachable in the default Layer 3 VRF on 10 switch models.

Affected models, fixes, and mitigations: https://thehackernews.com/2026/09/critical-cisco-nexus-9000-flaw-lets.html

🚨 This week’s ThreatsDay...

🧰 Real tools, wrong hands
🎣 Phishing for sale
🔗 Old access stays open
🤖 AI points to bad packages
⌨️ Typos open doors
🪪 ID data stolen
💥 Access becomes ransomware
⚙️ Weak settings remain

Full threat list: https://thehackernews.com/2026/09/threatsday-ceo-phishing-kits-5k-dropbox.html

⚡ GPT-6 Astra scored 100% on ExploitBench.

OpenAI says separate exploit-development tests included two zero-day vulnerabilities. The released version is limited to secure code review and patching and refuses requests to create PoC exploits.

Read: https://thehackernews.com/2026/09/gpt-6-astra-scores-100-on-exploitbench.html

‼️ WARNING - Attackers are exploiting a Chrome V8 zero-day.

CVE-2026-85046 allows arbitrary code execution inside the browser sandbox via a crafted HTML page. Google fixed it in Chrome 152.0.7977.82/.83.

How the V8 type confusion works: https://thehackernews.com/2026/09/google-releases-chrome-update-to-patch.html

⚠️ Plex is urging users to update after patching multiple undisclosed security flaws.

Fixes are available in Media Server 1.43.3 and Desktop 1.115.0. Plex has requested CVE IDs but disclosed no technical details.

What users need to update: https://thehackernews.com/2026/09/plex-urges-immediate-updates-after.html

🛑 Attackers are exploiting Super Forms and Elementor Pro RCE flaws.

Wordfence has blocked over 440,000 attempts. Both bugs allow unauthenticated PHP uploads; Elementor attacks require a published page with a Form widget and File Upload field.

Inside the bypass: https://thehackernews.com/2026/09/over-440000-exploit-attempts-target.html

🚨 Unauthenticated attackers could turn Grafana MCP servers into proxies for internal networks and cloud metadata endpoints.

Pillar Security research team found affected releases accepted self-generated session IDs in place of authentication, exposing the server's Grafana service account to anyone who could reach it.

A chained SSRF (CVE-2026-19516, CVSS 9.1) then let callers pick the destination, method, headers, and body of outbound requests. The server has passed 1.9 million Docker Hub downloads; Grafana shipped fixes in mcp-grafana v1.1.0.

Full details here: https://thn.news/grafana-mcp-flaws

🚨 Attackers compiled the newly documented Ted backdoor into HAProxy builds at two South Korean organizations.

It keeps C2 out of backend logs and HAProxy statistics while serving altered pages only to selected visitors.

How the filtering and log evasion work → https://thehackernews.com/2026/09/new-ted-backdoor-hides-inside-victims.html

‼️ PostgreSQL closed a 12-year-old flaw that could turn REPLICATION access into code execution.

With logical decoding enabled, an account could load any reachable library and run it as the postgres OS user.

How the flaw worked: https://thehackernews.com/2026/09/postgresql-fixes-12-year-old-logical.html