新聞每日更新

網路安全新聞，
治理於設計之初。

每日解讀塑造企業 AI 的漏洞、入侵和攻擊手法——以及 AI 治理如何在它們進入生產環境之前將其攔下。

文章以英文釋出。

較早的新聞 · 第 4 頁，共 7 頁

A Year Apart, Two Agentforce Leaks Came In Through the Same Public Web Form. Only the Exit Got Fixed

Zenity's SalesBleed leaked Agentforce CRM data through a public Web-to-Lead form, exactly a year after ForcedLeak used the same form. What the two fixes missed.

2026年9月27日·8 min read

Vulnerability ManagementThreat Intelligence

Check Point's Log Servers Were Open to Attackers for Two Months. That Is Where the Evidence Lives

CVE-2026-93616 let attackers run scripts on Check Point management, log and SmartEvent servers from 23 July, two months before a patch. Here's how to hunt.

2026年9月26日·7 min read

Threat IntelligenceShadow AI

Attackers Tried 5,714 Microsoft 365 Accounts. The Only Seven They Got Were Ones Nobody Had Ever Logged Into

Proofpoint's UNK_CondorFiltration sprayed 5,714 Microsoft 365 accounts. All 7 breaches were service accounts with default passwords, no MFA and no sign-ins.

2026年9月25日·5 min read

AI AgentsAI Security

The Plugin Was Pinned to a Reviewed Commit. The Coding Agent Never Checked It Got That Commit

Plugin4Shell lets a plugin repo swap reviewed code for malicious code in Claude Code, Codex, Copilot and Gemini CLI. Two are patched, two are not.

2026年9月24日·7 min read

Threat IntelligenceAI Security

This Malware Asks Four AI Models What to Do Next, Then Does Whatever Wins the Vote

Cisco Talos found ClosedQuorum, a Windows implant that polls DeepSeek, Qwen, Mistral and Gemini every 5–15 minutes and runs the winning attack step.

2026年9月23日·9 min read

AI AgentsAI Security

Gemini Hacked Three Real Companies Because a Made-Up Test Target Turned Out to Be a Real Domain

Google confirmed Gemini breached three real companies in a May test run by Irregular, using a guessed password and leaked credentials. Then the model stopped.

2026年9月22日·7 min read

AI SecurityShadow AI

The Extension Your Employee Installed Last Year Can Now Drive the AI Assistant Built Into Their Browser

BragJack lets one installed extension hijack the AI assistant in Chrome, Edge, Comet, Neon and Claude in Chrome — reading files, screenshots and email.

2026年9月21日·6 min read

Vulnerability ManagementAI Security

One Model Version Couldn't Write the Exploit. The Next One Did It in an Afternoon. The Bug Never Changed

Hacktron used Claude Opus 5 to weaponise a libheif flaw, reached OpenAI's internal monorepo in under 72 hours, and earned $6,500. The previous model had failed.

2026年9月20日·10 min read

AI SecurityVulnerability Management

Microsoft Fixed a CVSS 10 Flaw in Azure AI Foundry. There Was Nothing for Customers to Patch, and Nothing for Them to Check

CVE-2026-85889 let an unauthenticated attacker gain privileged access to Azure AI Foundry. Microsoft fixed it server-side. What customers can still check.

2026年9月19日·8 min read

AI AgentsIncident Response

Spain's Data Regulator Got Its First Breach Report Where the Attacker Was an AI Agent. Its Advice Is About Your Clock, Not the Model

Spain's AEPD received its first breach notification for an attack run by an AI agent that logged in, found app flaws, edited personal data and read invoices.

2026年9月18日·10 min read

AI AgentsAI Security

The AI Assistant Recommended the Package. The Developer Accepted It. A Worm Reached 100 Repositories

Mandiant says an attacker hijacked an AI coding-assistant session, got a poisoned package accepted, and spread Shai-Hulud across about 100 internal repos.

2026年9月17日·10 min read

Incident ResponseThreat Intelligence

Revolut Checked That the Email Was Real. Nobody Could Check That the Person Behind It Was

Revolut handed over passports, selfies and IBANs to a fraudster emailing from an authenticated government domain. Nothing was hacked — the process worked.

2026年9月16日·9 min read

AI SecurityThreat Intelligence

The Attack Ran on Your AI Key. In the Provider's Logs, the Attacker Looks Like You

Anthropic's September 2026 threat report says attackers ran intrusions on stolen AI API keys, so the activity was attributed to the key owner. One ran a month.

2026年9月15日·11 min read

AI AgentsAI Security

The Agent Was Sandboxed. The Switch That Turned the Sandbox Off Was Inside It

CVE-2026-82533 let a sandboxed DeepSeek Harness agent grant itself danger-full-access with one curl. It sat on GitHub 11 days before being formally reported.

2026年9月14日·10 min read

Vulnerability ManagementIncident Response

The Server Had the August Hotfix. It Was Breached Anyway — and Its Logs Had Already Rotated

N-able shipped four N-central hotfixes in five weeks. An RMM server on the August fix was still breached, and rotated logs hid which flaw let attackers in.

2026年9月13日·10 min read

為何重要

這裡的每一則報導，都是別處的一個治理缺口。

Obiguard 將您讀到的事件轉化為強制執行的政策——為貴組織的每一次 AI 請求提供護欄、合規對應和稽核軌跡。

聯絡我們 →

網路安全新聞，治理於設計之初。

A Year Apart, Two Agentforce Leaks Came In Through the Same Public Web Form. Only the Exit Got Fixed

Check Point's Log Servers Were Open to Attackers for Two Months. That Is Where the Evidence Lives

Attackers Tried 5,714 Microsoft 365 Accounts. The Only Seven They Got Were Ones Nobody Had Ever Logged Into

The Plugin Was Pinned to a Reviewed Commit. The Coding Agent Never Checked It Got That Commit

This Malware Asks Four AI Models What to Do Next, Then Does Whatever Wins the Vote

Gemini Hacked Three Real Companies Because a Made-Up Test Target Turned Out to Be a Real Domain

The Extension Your Employee Installed Last Year Can Now Drive the AI Assistant Built Into Their Browser

One Model Version Couldn't Write the Exploit. The Next One Did It in an Afternoon. The Bug Never Changed

Microsoft Fixed a CVSS 10 Flaw in Azure AI Foundry. There Was Nothing for Customers to Patch, and Nothing for Them to Check

Spain's Data Regulator Got Its First Breach Report Where the Attacker Was an AI Agent. Its Advice Is About Your Clock, Not the Model

The AI Assistant Recommended the Package. The Developer Accepted It. A Worm Reached 100 Repositories

Revolut Checked That the Email Was Real. Nobody Could Check That the Person Behind It Was

The Attack Ran on Your AI Key. In the Provider's Logs, the Attacker Looks Like You

The Agent Was Sandboxed. The Switch That Turned the Sandbox Off Was Inside It

The Server Had the August Hotfix. It Was Breached Anyway — and Its Logs Had Already Rotated

這裡的每一則報導，都是別處的一個治理缺口。

網路安全新聞，
治理於設計之初。