Applying Security Engineering to Prompt Injection Security

Applying Security Engineering to Prompt Injection Security

This seems like an important advance in LLM security against prompt injection:
Google DeepMind has unveiled CaMeL (CApabilities for MachinE Learning), a new approach to stopping prompt-injection attacks that abandons the failed strategy of having AI models police themselves. Instead, CaMeL treats language models as fundamentally untrusted components within a secure software framework, creating clear boundaries between user commands and potentially malicious content.
[…]
To understand CaMeL, you need to understand that prompt injections happen when AI systems can’t distinguish between legitimate user commands and malicious instructions hidden in content they’re processing…

Read More
Product Walkthrough: Securing Microsoft Copilot with Reco

Product Walkthrough: Securing Microsoft Copilot with Reco

Find out how Reco keeps Microsoft 365 Copilot safe by spotting risky prompts, protecting data, managing user access, and identifying threats – all while keeping productivity high.

Microsoft 365 Copilot promises to boost productivity by turning natural language prompts into actions. Employees can generate reports, comb through data, or get instant answers just by asking Copilot. 
However,

Read More
Google Reports 75 Zero-Days Exploited in 2024 — 44% Targeted Enterprise Security Products

Google Reports 75 Zero-Days Exploited in 2024 — 44% Targeted Enterprise Security Products

Google has revealed that it observed 75 zero-day vulnerabilities exploited in the wild in 2024, down from 98 in 2023. 
Of the 75 zero-days, 44% of them targeted enterprise products. As many as 20 flaws were identified in security software and appliances.
“Zero-day exploitation of browsers and mobile devices fell drastically, decreasing by about a third for browsers and by about half for

Read More
Malware Attack Targets World Uyghur Congress Leaders via Trojanized UyghurEdit++ Tool

Malware Attack Targets World Uyghur Congress Leaders via Trojanized UyghurEdit++ Tool

In a new campaign detected in March 2025, senior members of the World Uyghur Congress (WUC) living in exile have been targeted by a Windows-based malware that’s capable of conducting surveillance.
The spear-phishing campaign involved the use of a trojanized version of a legitimate open-source word processing and spell check tool called UyghurEdit++ developed to support the use of the Uyghur

Read More
CISA Adds Actively Exploited Broadcom and Commvault Flaws to KEV Database

CISA Adds Actively Exploited Broadcom and Commvault Flaws to KEV Database

The U.S. Cybersecurity and Infrastructure Security Agency (CISA) on Monday added two high-severity security flaws impacting Broadcom Brocade Fabric OS and Commvault Web Server to its Known Exploited Vulnerabilities (KEV) catalog, citing evidence of active exploitation in the wild.
The vulnerabilities in question are listed below –

CVE-2025-1976 (CVSS score: 8.6) – A code injection flaw

Read More
X