Security | Threat Detection | Cyberattacks | DevSecOps | Compliance

AI has changed the calculus of a credential attack. Before, finding and exploiting credentials in an enterprise environment required time, patience, and human judgment. An attacker had to decide which accounts were worth testing and which systems were worth reaching. Many credentials never made the list.

We studied what happens when Large Language Models (LLMs) generate vulnerability patches for recently disclosed, complex vulnerabilities. Our data shows that LLMs produce Fix-Like Artifacts with Embedded Defects (FLAWED) 53.9% of the time when complex patches are required.

It's 6:30am and you hear the door of the nightclub you've spent the last 8 hours inside shriek as it closes behind you. You watch bleary-eyed as an overly bright sunrise illuminates the business-suited people as they glide effortlessly along the sidewalk, their obnoxiously well-rested faces talking about work on their fully charged phones. You wonder, "Where did all the fun people go? And what happened to my wallet?".

Ask most security tools what an AI coding agent just built, and they can tell you. Ask what it was allowed to consume before it started, and far fewer have an answer. That gap, between watching agentic development and controlling it, is what securing it actually comes down to. Securing agentic development means stopping risk before it enters a build, not flagging it after. And risk prevention has a prerequisite most approaches skip: you can only account for the assets you actually hold and manage.

Today, we’re excited to announce a few developer experience improvements to Cloudflare AI Search to make it easy to manage a search solution out of the box. Previously, you had to stitch together components of the Cloudflare primitives (Workers AI, AI Gateway, Vectorize, R2, Browser Run) but now, AI Search can do this automatically, and better. Our goal is to give your agents their own search engine, where they can easily find data to provide better answers for themselves and their humans.

Enterprise AI guardrails are the technical controls that prevent AI systems from doing things they shouldn't, applied at the moment of execution rather than after the fact. They sit between the AI model or agent and the systems, data, and users it interacts with, filtering inputs, inspecting outputs, and constraining behavior against enterprise policy.

Monitoring AI agents in production is a fundamentally different problem from monitoring traditional software or even generative AI models. Because agents run autonomously, chain multi-step reasoning across tools and systems, and change behavior as their underlying models evolve, standard software metrics like uptime and CPU utilization miss almost everything that matters. ‍

Every SaaS tool your organization onboards now carries a hidden layer of AI risk. The chatbot on your CRM, the transcription service your sales team runs, the code assistant embedded in your IDE. Each one processes company data through models you did not build, in ways your vendor questionnaire was not written to catch. Traditional third-party risk management was designed to evaluate infrastructure, access controls, and data handling.

Most guides tell security teams to pick an access control model and move on. That advice breaks down in real environments, because the hard part isn't naming the model, it's keeping least privilege, auditability, and enforcement aligned as identities, attributes, and relationships keep changing across SaaS, cloud, endpoints, and network gear. In practice, the winners are the teams that treat access control as an operating discipline, not a diagram in an architecture deck.

Most of the exposure research we publish goes one way. We count some category of internet-exposed devices, the number is bigger than you hoped, and it is climbing. Webcams. Industrial control systems. NTP servers old enough to vote. The story is almost always "there is more of this exposed than there should be, and it is getting worse.".