🚨 TerminalFix turns a fake Cloudflare CAPTCHA into a reverse tunnel inside the victim’s network.

Microsoft says the ClickFix variant tricks users into running PowerShell, then deploys a backdoor that can route traffic to other hosts visible from the infected machine.

Read: https://thehackernews.com/2026/08/terminalfix-uses-fake-cloudflare.html

DoJ corrected its claim that several U.S. agencies were QTFY victims.

NASA, the Fed, DOE, DOJ, HHS, NIH and the U.S. Senate are now listed as targets instead. The revised statement does not establish that those agencies were compromised.

Read: https://thehackernews.com/2026/08/doj-corrects-china-hacking-claim-says.html

🚨 Fire Ant turned compromised Cisco routers into covert traffic and credential collection points.

The actor captured packet data, harvested TACACS credentials, and suppressed router logs and telemetry while probing connected high-value networks.

Read: https://thehackernews.com/2026/08/china-linked-fire-ant-hijacks-cisco.html

⚠️ Aurora ransomware operators gave Cursor AI credentials or an existing route into victim networks.

The agent handled exploitation tasks against 10 targets, including network scanning, privilege checks, NTLM relay attempts, and certificate attacks.

Read: https://thehackernews.com/2026/08/aurora-ransomware-operators-use-cursor.html

Anthropic’s new Claude Code transcripts show actions, not whether access is legitimate.

They capture prompts, bash commands, file operations, and MCP calls, but miss local hooks, unused configs, and sessions using non-Anthropic models.

Where visibility breaks: https://thehackernews.com/2026/08/securing-claude-code-new-compliance-api.html

⚡ Trust did the damage this week.

• AI agents breached Hugging Face
• FBI disrupted Chinese spy proxies
• Backdoors shipped in routers
• PaperCut flaws came under attack
• Fake IT support pushed ransomware
• Malware targeted 1,243+ banking apps

Full recap: https://thehackernews.com/2026/08/weekly-recap-chinese-spy-proxy-ai.html

ValleyRAT runs inside a signed wallpaper app to evade security controls.

The modified QN Wallpaper installer disables Windows Defender, then sideloads the backdoor under QnWallpaper.exe. ValleyRAT can capture keystrokes, clipboard contents and screenshots.

How the disguise works: https://thehackernews.com/2026/08/valleyrat-backdoor-hides-in-signed.html

⚠️ Attackers are exploiting critical Langflow and Ruby on Rails flaws.

Langflow attacks probe OpenAI and AWS credentials. The Rails bug can leak secrets and lead to RCE in apps processing untrusted uploads with libvips; observed activity established C2.

How the attacks work: https://thehackernews.com/2026/09/attackers-exploit-critical-langflow-and.html

AI coding agents can change code they do not fully understand.

HoundDog AI co-founder Amjad Afanah reports that, in one test, a one-line protobuf change affected four services, three teams, and two languages.

Why bigger context windows are not enough: https://thehackernews.com/expert-insights/2026/08/the-missing-context-layer-for-ai-agents.html

🚨 Attackers stole a METR API key and used it for three weeks, consuming credits worth about $600,000.

A fail-open bug disabled Google authentication on a public agent dashboard. The attacker prompted an agent to reveal the key and added SSH persistence.

How the exposed app was abused: https://thehackernews.com/2026/09/attackers-steal-metr-api-key-and.html

Iranian hackers are posing as recruiters on LinkedIn and hiding cross-platform remote access trojans inside coding tests.

The campaign, attributed to Nimbus Manticore, delivers two newly documented malware families: NodeRabbit and PollCat. Both can execute commands and establish persistence across Windows, Linux, and macOS.

Read: https://thehackernews.com/2026/09/iranian-hackers-pose-as-recruiters-to.html

Russia-aligned hackers planted a nuclear weapon prompt in malware to stop AI analysis.

UAC-0099 used the malicious VBS script against a target in Ukraine. It is designed to install MATCHBOIL, the actor’s C# loader.

Inside the GuardBreaker anti-analysis trick: https://thehackernews.com/2026/09/russia-aligned-uac-0099-plants-nuclear.html

‼️ 13 malicious Packagist packages target unpatched iPhones with wallet-stealing spyware.

Installed on Vietnamese streaming sites, the themes serve a WebKit-to-kernel chain that can exfiltrate passwords, messages, photos, location data, and crypto wallet seeds.

Read: https://thehackernews.com/2026/09/13-malicious-packagist-packages-target.html

🚨 Breeze Comet used compromised accounts for hundreds of fraudulent transactions.

The Brazil-based group deploys COBALTSPIN to reach core financial apps and payment APIs. At least one heist netted tens of thousands of dollars.

Read: https://thehackernews.com/2026/09/breeze-comet-executes-hundreds-of.html

ClickFix accounted for 47% of attacks cited in Microsoft security notifications.

Built-in tools appeared in 84% of Bitdefender’s high-severity incidents. Novelty is not the advantage. Repeatability is.

Inside the attack playbooks that keep working: https://thehackernews.com/2026/09/threat-actors-dont-want-better-attacks.html

🛑 Attackers exploit a JFrog Artifactory flaw to mint admin tokens.

The critical auth bypass affects default configurations, requires network access but no login, and was weaponized days after disclosure.

How the attack works: https://thehackernews.com/2026/09/attackers-exploit-critical-jfrog.html

⚡ Authorities turned Sality’s own P2P network against the botnet.

Peer-list manipulation redirected more than 15,000 infected machines to CrowdStrike sinkholes, blocking new payloads. Existing malware remains active and must be removed.

How the botnet was trapped: https://thehackernews.com/2026/09/authorities-turn-salitys-p2p-network.html

🛑 Attackers are exploiting Switchvox without credentials to deploy reverse shells.

CVE-2026-9586 is a critical SQL injection in the /pa endpoint. About 4,000 instances are exposed online, and Sangoma has released a patch.

Inside the exploit: https://thehackernews.com/2026/09/attackers-exploit-critical-switchvox.html

Shadow AI is hiding inside approved AI tools.

Akto CEO Ankita Gupta cites roughly 7,600 malicious GitHub repos, including 800+ posing as AI Skills or MCP servers. The app is approved. Its hidden capabilities may not be.

Inside the agent supply-chain gap:
https://thehackernews.com/expert-insights/2026/08/shadow-ai-is-now-hiding-inside.html

‼️ Researchers used Claude to port a pre-auth RCE exploit from one PLC model to another. It ran ARM shellcode on live hardware.

A later C2 implant attempt wrote to flash-mapped memory and permanently bricked the target.

How the port worked: https://thehackernews.com/2026/09/researchers-use-claude-to-port-pre-auth.html