Security | Threat Detection | Cyberattacks | DevSecOps | Compliance

As large language models (LLMs) become more embedded in business operations, the risks and attack methods targeting them are evolving just as quickly. The 2025 edition of the OWASP Top 10 for LLM Applications reflects this rapid evolution, addressing the current threats facing generative AI systems in production environments. For organizations investing in LLMs, understanding the risks is crucial for deploying these systems securely.

The NetSuite AI Connector Service enables external AI agents to authenticate directly into NetSuite using real user identities and MCP-based tool execution. While Oracle limits elevated actions at the platform level, AI agents still inherit the full permission scope of the connected role. That shifts longstanding governance weaknesses, including over-permissioned roles, SoD conflicts, and undocumented customizations, into active operational risk.

Cybersecurity operations were already becoming increasingly difficult to scale long before AI-driven and increasingly agentic attacks began accelerating the threat landscape. Customer environments continued expanding across endpoints, identities, cloud services, SaaS applications, remote users, and operational infrastructure. More environments created more telemetry, more coordination, and more operational complexity for teams already operating near capacity.

This week, Connie Loizos, editor in chief of TechCrunch, sat down backstage with Francis de Souza, COO of Google Cloud, for a piece on the state of enterprise AI security. The interview is worth reading in full. Three points in it should reshape how every CISO is thinking about the next twelve months.

A few weeks ago I was on a call with a financial services customer who had moved a credit-decisioning model into production on Red Hat OpenShift AI. They were happy with the platform. They were less happy with the answer they had for a question their risk officer had just asked: “If an attacker encrypts the cluster tomorrow, what do we need to bring back to be inference-ready by Monday morning?” The team started listing the obvious things — the model artifact, the serving endpoint.

With over 10 years of experience in implementing AI, KnowBe4 has a ton of agents on its platform which customers can use to significantly lower risk. They help to secure the digital workforce (humans + AI agents). But five of them, all based and driven by risk scoring metrics, have become my favorites.

Download Beyond Identity: The CISO's Guide to Securing Agentic AI for a 12-month roadmap to comprehensive agent governance, starting with visibility. Some organizations still treat agentic AI as a future problem. Something to plan for. Something on the horizon. That framing is wrong, and the inaction it entails will put you behind.

Every conversation I have with security leaders about enterprise AI security eventually arrives at the same place: a description of what they've extended. Their data loss prevention tool now flags sensitive data going into prompts. Their SIEM is ingesting AI platform logs. Their cloud security team has added model endpoints to their coverage scope. For many teams, this represents real effort and real progress.