‼️ You can commit code to a GitLab repo by emailing it.

Every GitLab user gets a private address for filing issues. Hidden inside is a non-expiring account token. If that address leaks, someone can send a patch that GitLab commits as you and, with your permissions, target main or trigger CI/CD.

Incoming email also bypasses 2FA and IP restrictions.

Read: https://thehackernews.com/2026/09/a-leaked-gitlab-issue-email-address.html

Attackers are using malicious Terraform providers to deliver Go malware through HashiCorp’s registry.

The malware uses Ethereum smart contracts and Slack as dual C2 channels and can execute Go or JavaScript commands.

Inside the supply-chain technique: https://thehackernews.com/2026/09/attackers-use-malicious-terraform.html

⚡ Attackers are exploiting WordPress CVE-2026-87902 within hours of disclosure.

The attacks use pearcmd.php to write attacker-controlled PHP files to disk, but only when specific theme and server conditions are met.

🔗 How the exploit works → https://thehackernews.com/2026/09/attackers-exploit-wordpress-cve-2026.html

⚠️ TeamFiltration targeted 5,700+ Microsoft 365 accounts. Seven were compromised, all unmonitored service accounts still using default passwords.

The campaign spanned 28 tenants and used password spraying against dormant identities.

Learn more about this attack → https://thehackernews.com/2026/09/teamfiltration-compromises-seven.html

Samsung’s mobile AI security model starts with where data goes.

The enterprise playbook: keep sensitive tasks on-device when possible, govern cloud processing centrally, tie AI to managed corporate identities, and build it on a trusted device foundation. Galaxy AI and Knox are Samsung’s example.

The four questions: https://thehackernews.com/expert-insights/2026/09/enterprise-mobile-ai-security-trade.html

🚨 OpenAI’s AI agent bypassed controls on an Australian Medicare statistics portal and accessed non-public files.

PM Anthony Albanese said OpenAI took too long to notify the government and called its disclosure unacceptable.

🔗 Here's what happened → https://thehackernews.com/2026/09/openai-agent-bypassed-australian.html

Your AI productivity gain has a hidden headcount cost.

ActiveState's new report, The AI Remediation Bill, puts a number on AI's downstream cost: engineers spend a conservative floor of 6.7% of their time fixing or revisiting AI-assisted development, about 139 hours a year, which for a 500-engineer org is nearly 34 full-time engineers.

Join the waitlist → https://thn.news/ai-remediation-bill

A fake Cloudflare check ended with Vidar Stealer running inside a signed Microsoft process.

Researchers mapped 17,000+ infected URLs. About 3,000 still served ClickFix lures, while Polygon smart contracts let operators swap lure hosts without touching the compromised sites.

Inside the chain: https://thehackernews.com/2026/09/17000-urls-reveal-how-clickfix-turns.html

🚨 Attackers use hacked Ukrainian sites to serve fake Cloudflare CAPTCHAs for Psychedelic Stealer.

The ClickFix lure copies an msiexec command and tells users to run it. The stealer targets passwords, tokens, and crypto-wallet data.

🔗 Read → https://thehackernews.com/2026/09/hacked-ukrainian-sites-serve-fake.html

Every unpatched CVE is a bet that no one exploits it before you fix it. And for teams stuck on older versions of Java, the odds get worse every year you put off patching. We’re talking 242 new CVEs published daily, up 80% from 2025

If you're stuck on hard-to-upgrade Java versions, pulling a backported version keeps your application secure while you focus on shipping what’s next. Chainguard Libraries for Java now includes CVE remediation, letting you stay safe while you plan your next major upgrade.

Learn More → https://thn.news/cve-remediation

⚠️ AI-assisted commits leak secrets at roughly twice the rate of human-written ones.

Coding agents can read .env files and MCP configs, while the same credentials may persist across CI/CD, tickets, and collaboration tools.

🔗 How the sprawl happens → https://thehackernews.com/2026/09/secrets-sprawl-is-identity-problem-that.html

Corp MDM spyware steals new SMS and can redirect calls on Android devices.

The logistics-focused campaign delivers it through fake Google Play pages. Once sideloaded and granted SMS and phone permissions, the app hides its launcher and sends new inbound texts to attacker infrastructure.

How it works: https://thehackernews.com/2026/09/corp-mdm-spyware-targets-logistics.html

‼️third-party[.]com is now serving Windows users a ClickFix lure disguised as a Cloudflare check.

The domain is referenced as a placeholder across 1,700+ repositories. The lure copies a PowerShell command to the clipboard, while non-Windows visitors see a decoy.

🔗 here’s how it works → https://thehackernews.com/2026/09/placeholder-third-partycom-referenced.html

‼️ The attacks worth watching this week barely look like attacks.

>> A search result
>> A coding tool
>> A software update
>> One normal-looking link

Behind them: AI search poisoning, leaked source code, malicious updates, supply-chain attacks, EDR evasion, and one-click code execution.

16 stories in this week’s ThreatsDay: https://thehackernews.com/2026/09/threatsday-ai-search-poisoning-ai.html

‼️ OnePlus confirmed a stock phone could be rooted by an installed app. Then it warned the researcher not to publish.

The app needs no special permissions. OnePlus confirmed the two flaws in May, but no fix was available when the researcher disclosed them. No real-world exploitation is known.

How the root chain works → https://thehackernews.com/2026/09/unpatched-oneplus-flaws-let-installed.html

WSO2 and Adobe Commerce flaws are being exploited in attacks.

The WSO2 bug can lead to remote code execution through unrestricted file upload. The Adobe flaw can switch a customer session to another account. CISA has added both to KEV, with federal agencies given until Sept. 27 to patch.

What attackers can exploit: https://thehackernews.com/2026/09/wso2-and-adobe-commerce-flaws-exploited.html

🚨 Cloudflare Containers flaw let customers read other tenants' leftover databases and credential files.

Researchers recovered residual data on 18 of 24 production placements by reading reused disk blocks. Cloudflare fixed the flaw and says it found no evidence of abuse beyond authorized testing.

How it worked → https://thehackernews.com/2026/09/cloudflare-fixes-flaw-that-let-one.html

🚨 Attackers are exploiting a Roundcube pre-auth SQL injection flaw patched in May.

CVE-2026-48842 can expose mail credentials and stored messages through the virtuser_query plugin.

Read: https://thehackernews.com/2026/09/roundcube-pre-auth-sql-injection-flaw.html

Bitget says suspected North Korean hackers stole $351.6 million after compromising a backend wallet system.

The attackers spoofed transaction data and triggered Bitget’s authorization process to move funds. Withdrawals are temporarily suspended while the breach is investigated.

What Bitget says happened: https://thehackernews.com/2026/09/bitget-says-suspected-north-korean.html

AI can turn a failed privilege-escalation attempt into another test within minutes.

Attackers get faster troubleshooting and script fixes. SOC teams still lose evidence, confidence, and context across handoffs, forcing analysts to rebuild the same incident.

That speed gap is the real problem → https://thehackernews.com/2026/09/the-soc-doesnt-need-to-start-over-with.html