️ /Cybersecurity Brief [Friday, February 20, 2026 | 00:54 UTC Edition]: AI

BLUF: This intelligence cycle reveals a significant shift in the threat landscape with the emergence of PromptSpy, the first Android malware leveraging generative AI for persistence. Major tech players are also under scrutiny, as a critical Microsoft Copilot bug exposed confidential emails and a severe RCE flaw was found in Microsoft's AI SDK. Further concerns arise from allegations of a biometric data firm sharing user IDs with governments and the national security implications of CISA's operational limitations due to a DHS shutdown.

📰 INDUSTRY INTEL (The Big 5)

PromptSpy: First Android Malware Leverages Generative AI for Persistence

The Scoop: Researchers discovered PromptSpy, the first Android malware utilizing Google's Gemini AI chatbot to automate its execution flow and achieve persistence on devices.

Why It Matters: This marks a new era in mobile malware, demonstrating AI's potential for more sophisticated and evasive attacks, raising concerns for future AI-powered threats.

Link:

Microsoft Copilot Bug Exposed Confidential Customer Emails for Weeks

The Scoop: A flaw in Microsoft Copilot mistakenly allowed the AI tool to read and summarize confidential customer emails for weeks before an update was rolled out to fix the issue.

Why It Matters: This incident highlights the critical data privacy and security risks associated with integrating generative AI into enterprise applications, even from major vendors.

Link:

Persona Accused of Secretly Storing Biometrics, Sharing with Governments

The Scoop: Cybersecurity researchers allege Persona, an age verification service, is retaining user biometrics and government photo IDs for three years and sharing them with US and Canadian federal agencies.

Why It Matters: These allegations raise serious privacy concerns biometric data handling and potential government surveillance through private companies.

Link:

Critical RCE Flaw Found in Microsoft's Semantic Kernel Python SDK

The Scoop: Microsoft's Semantic Kernel Python SDK, a key component for building AI applications, contains a critical remote code execution vulnerability (CVE-2026-26030, CVSS 9.9).

Why It Matters: This RCE poses a severe threat to applications leveraging the SDK, potentially allowing attackers to execute arbitrary code and compromise AI-driven systems.

Link:

DHS Partial Shutdown Significantly Impacts CISA's Cyber Defense Capabilities

The Scoop: A partial DHS shutdown, ongoing since February 14, 2026, due to a lapse in appropriations, has severely impacted the Cybersecurity and Infrastructure Security Agency (CISA).

Why It Matters: A weakened CISA during a shutdown could leave critical infrastructure and civilian networks more vulnerable to cyberattacks, raising national security concerns.

Link:

* - Critical RCE in Velocity Template Engine (CVSS 10.0) -

* - Microsoft Semantic Kernel SDK Vulnerable to Critical RCE (CVSS 9.9) -

* - Hyland Alfresco Transformation Service Exposed to Critical RCE (CVSS 9.8) -

* - Gogs Self-Hosted Git Service Suffers Critical Unauthenticated Access Vulnerability (CVSS 9.8) -

* - RustFly 2.0.0 Contains Critical Command Injection Vulnerability (CVSS 9.8) -

* - PromptSpy: First Android Malware Leveraging Gemini AI for Runtime Persistence Discovered -

* - Microsoft Copilot Bug Exposed Confidential Customer Emails for Weeks -

* - Peter Thiel-Funded Persona Accused of Storing Biometrics and Sharing with US/Canada Gov -

* - Supply Chain Attack Secretly Installs OpenClaw for Cline Users -

* - University of Mississippi Medical Center Hit by Cyberattack, Forcing Clinic Closures -

* - FBI Warns of Rising ATM 'Jackpotting' Attacks Netting Millions -

* - Texas Sues TP-Link Over Alleged Chinese Hacking Risks and User Deception -

🛡️ Patch Watch (Top 10)

* - Velocity Template Engine: Remote Code Execution (CVE-2025-12107) - [Link]( )

* - Semantic Kernel: Remote Code Execution vulnerability (CVE-2026-26030) - [Link]( )

* - RustFly 2.0.0: Command injection in remote UI control mechanism (CVE-2026-27476) - [Link]( )

* - Hyland Alfresco Transformation Service: Unauthenticated Remote Code Execution (CVE-2026-26339) - [Link]( )

* - Clasifico Listing plugin for WordPress: Privilege Escalation (CVE-2025-12882) - [Link]( )

* - Slider Future plugin for WordPress: Arbitrary File Uploads (CVE-2026-1405) - [Link]( )

* - Prodigy Commerce plugin for WordPress: Local File Inclusion (CVE-2026-0926) - [Link]( )

* - s2Member plugin for WordPress: Privilege Escalation via account takeover (CVE-2026-1994) - [Link]( )

* - 'Saisies pour formulaire' (Saisies) plugin for SPIP: Remote Code Execution (CVE-2025-71243) - [Link]( )

* - Inradius: Execution After Redirect (EAR) and Missing Authentication for Critical Function (CVE-2025-8350) - [Link]( )

🧠 Intelligence Cycle Question:

With the rise of AI being leveraged in both offensive (e.g., PromptSpy malware) and defensive cybersecurity strategies, how do you foresee this dual-use technology reshaping the threat landscape and the skills required for security professionals in the 5 years?