Skip to content

Sam's News — security — 2026-09-13

Security

7.5 Microsoft cloud accounts hijacked through passkey phishing and email fraud

Microsoft disclosed that threat actors are using passkey-themed social engineering and third-party email infrastructure to breach cloud accounts and steal data.

  • Active since May 2026
  • Attacks begin with phone/message impersonation of IT helpdesk
  • Identity compromise enables cloud reconnaissance and data exfiltration
  • SharePoint, OneDrive, email collection via REST APIs
  • Proxy-associated infrastructure used for automated collection

Sources: The Hacker News RSS Update to: Passkey-themed social engineering enables identity and cloud compromise attacks

7 China-linked hackers exploit Tencent flaw to deploy GrayRabbit malware

Threat actors associated with a China-aligned espionage group are exploiting CVE-2026-51990 in Tencent's Sogou Input Method to deploy the GrayRabbit backdoor.

  • Sogou Input Method: 455+ million monthly users across Windows, Android, iOS; ~70% of Chinese input-method market share
  • Attack vector: malicious sgbiz: protocol link passed to biz_helper.exe without argument validation
  • Exploit chain used browser rendering of malicious page via Sogou's skin store; Chromium version 80 (from 2020) had sandbox and same-origin policy disabled
  • Tencent patched vulnerability April 2026 but only blocked attack vector, not underlying browser engine issues

Sources: BleepingComputer RSS Update to: China-linked hackers exploited Sogou Input Method flaw to deploy GRAYRABBIT backdoor

6.5 NSA Undergoes Major Reorganization Creating Five Mission Centers

The National Security Agency is implementing a comprehensive reorganization that will establish five mission centers, including dedicated cyber and AI divisions.

Sources: The Record RSS

6.5 Mathspace Security Breach Reveals 23-Day Patch Gap in School Systems

Educational platform Mathspace suffers a security breach that exposed a significant delay in applying critical security patches.

Sources: The Educator K/12 RSS

AI

7.5 Anthropic CEO warns AI could achieve dangerous autonomy within 6-12 months

Anthropic CEO Dario Amodei warned on September 13, 2026 that AI could achieve dangerous autonomy within 6–12 months, capable of commanding swarms of autonomous agents that could compromise internet infrastructure. He called for the AI industry to slow development to allow safety measures to advance in parallel.

  • AI could take over entire internet within 6–12 months, Amodei warned
  • Slowing development by extra year or two could significantly reduce serious harm risk
  • Amodei said coordination required across AI industry, governments, and authoritarian regimes
  • Anthropic implementing safety measures independently; other components need broad coordination
  • OpenAI CEO stated company delaying Wall Street stock sale to prioritize safety

Sources: SecurityWeek AI Web Searched