Security | Threat Detection | Cyberattacks | DevSecOps | Compliance

If you’ve been in security operations for more than a few years, you’ve lived through the automation hype cycle at least twice. First, it was SIEM that was going to solve everything. Then SOAR was supposed to fix what SIEM couldn’t. Now, AI SOC platforms are delivering what SOAR always promised but never actually could.

A blanket “no AI” policy is not the answer for most organizations. Prohibiting something people find genuinely useful just drives it further underground.

Most security teams are being asked to "enable AI" before they have any real sense of which tools are safe to use. That gap is costing them. Cyberhaven's research found that the majority of AI tools in active enterprise use today fall into high or critical risk categories, and more than 80% of enterprise data flowing into AI is going to those risky tools, not to platforms built with serious security in mind. To help security teams cut through the noise, we built the Cyberhaven AI App Risk Checker.

Today, we’re launching System Prompt Hardening, Mend.io’s new capability that defends the hidden instructions that control how your AI systems behave. Unlike user-facing prompts, system prompts live behind the scenes, and when attackers manipulate them, the result can be data leaks, policy bypasses, or unsafe model behavior. System prompt hardening stops those attacks at the source and gives security, engineering, and risk teams a practical, auditable way to secure AI in production.

Most security teams cannot confidently answer a simple question: who has access to which cloud resources right now? Human identities and accounts now span across thousands of services, subscriptions, and SaaS platforms. The result is a vast, decentralized environment riddled with “unknown unknowns” that security teams cannot fully map, and that traditional security controls weren’t designed to address. Attackers count on these identity blind spots.

If your identity governance program feels like a relic from a simpler time, you’re not alone. Traditional identity governance and automation (IGA) was built for a world where job titles told the whole story. A software engineer was a software engineer; a sales rep was a sales rep. Assigning access was intended to be as simple as slotting people into predefined roles.

When cybersecurity teams talk about risk, they usually speak in technical terms like vulnerabilities, exploits, and attack vectors. But when they walk into the boardroom, they need to speak a different language. They need to speak about cost. In the era of AI, the cost of insecure APIs has shifted from a potential liability to a tangible line item on the balance sheet. It is no longer just about the cost of a data breach.

Long-time followers of mine know that I am not an AI hype person. Some people might even call me an AI critic. I prefer to call myself an AI realist. I do not think AI will kill us all (despite our best efforts to bypass all guardrails and common sense). I do not think AI will replace all jobs. I do not think AI will replace all cybersecurity jobs. But I do think AI allows improvements in many areas, including cyber defenses, over traditional tools and techniques.

Production deployments. That’s where multi-agent AI systems live now, not research labs. Salesforce, Microsoft, and Cognition Labs are all running agent pipelines that replaced what used to take entire ops teams. Most businesses still don’t fully understand what they’ve switched on. A multi-agent AI setup isn’t just one model doing more things.