Cybersecurity

TL;DR Prompt injection is an attack where untrusted text, either typed by a user or hidden in content an AI application reads, is treated by a large language model (LLM) as instructions. This can make the model ignore its rules or take actions it shouldn’t. Prompt injection is ranked the...

If you’re trying to work out which Microsoft Purview licence you actually need (and what it’s going to cost you if you get it wrong), you’re not alone. It’s one of the most common questions our team fields, and the answer is rarely as straightforward as Microsoft’s licence table makes...

AI Cybersecurity: 8 Things Your IT Teams Need to Know In 2026 AI is changing how attackers work and how organisations manage risk. When deciding how your organisation should embrace AI, cybersecurity should be top of the consideration list. For IT leaders, a priority is control of AI tools that...

Microsoft Purview Licensing Explained: Business Premium vs E3 vs E5 If you’ve looked into Microsoft Purview and come away confused about which license you actually need, you’re not alone. It’s the single biggest blocker CloudGuard sees when SMBs and mid-sized organisations start a data governance project, not the technology, the...

Why Multi-Model Security Changes Everything  Six years building an agentic SOC analyst (ANSEL) teaches you something quickly: more data is critical but not the answer. Better understanding through context of what it means is.   Microsoft Project Perception is built on exactly that insight. It’s not another security product. It’s a different way of thinking about how AI should reason, with context, consequence, and...

Most vendor evaluation processes are built to survive procurement, not to protect you eighteen months after go-live. Here’s the gap almost nobody catches before signing. Outlining The Problem The majority of security technologies need 90 days just to establish an accurate behavioural baseline and fair comparison. Please remember your existing...

We’ve all done the training, so why are attackers still getting through? Attackers no longer rely on bad spelling or suspicious links, they use AI-generated deepfakes and psychological profiling to manipulate people with astonishing precision. By exploiting the brain’s emergency response system, they trigger fear, urgency, or authority to override...

When I shared my last article, a few people got in touch asking for a more practical follow-up, specifically around how small teams can use Claude Pro without putting business data at risk. This piece goes step by step through exactly that. Understand what you’re actually adopting Claude Pro is...

If your organisation thinks a password reset or MFA alone are enough, think again. In this phishing attack breakdown by CloudGuard’s SOC team, Conor and Jon reveal the reality behind an actual breach involving a UK law firm, exposing how hackers use four methods to regain access long after initial...