MCP Authorization: What Belongs on the Wire (and What Doesn't)
MCP makes it remarkably easy to connect AI agents to external tools. Securing those connections is another matter.
Posts from every feed QUEST has found across the web — blogs, news, podcasts, channels and more.
NDC Security 2027 is a 4-Day Event for Software Developers in Oslo, with a focus on Security. 15-18 February 2027 - Radisson Blu Scandinavia Hotel.
MCP makes it remarkably easy to connect AI agents to external tools. Securing those connections is another matter.
We all (mostly) love AI. Those wonderful agents spew out all the glorious code we ask for, and they just keep getting better at it. Productivity is through the roof! We’re shipping new things left, right and center!
I run an AI Battle Arena at arena.angel-serv.com. You plug your CLI in (Claude Code, Codex, Gemini, openclaw) and fight 16 bots on a world leaderboard. We want you to try to break the server.
At this point, agents are everywhere and they are pretty hard to ignore. They are showing up in our CI/CD life cycles, in code development, in code reviews, and in the day-to-day workflows that engineering teams are encouraged to adopt by both lower and upper management.
There is a common saying that "every problem in cryptography can be reduced to key management problem". What if we could make life easier for us in this area? TPMs (Trusted Platform Modules) have been a fixed part of every standard PC for many years, providing all users with a "free" hardware that can be used for all kinds of cryptography.
What happens when an agent is trusted to communicate with your system, but its messages are trusted to control your system?
In March 2026, while reverse-engineering the cloud platform behind a popular line of consumer baby monitors and home security cameras, I discovered that one MQTT SUBSCRIBE wildcard returned the live message stream from every device on the platform.
I run an AI Battle Arena at arena.angel-serv.com. You plug your CLI in (Claude Code, Codex, Gemini, openclaw) and fight 16 bots on a world leaderboard. We want you to try to break the server.
AI orchestration platforms promise to automate your life. They deliver, just not always for you. Kestra, Langflow, Nocobase, Flowise, Activepieces, Dify, and Apache Airflow have quietly become critical infrastructure, and they all share the same dangerous assumption: anyone who can touch a workflow is trusted to run code on the host.
From December 2026, every EU Member State must offer its citizens a Digital Identity Wallet (EUDI Wallet). This represents the largest deployment of privacy-preserving credential technology to date. Hundreds of millions of people will then be able to prove their identity, that they are over 18, or that they hold a diploma, directly from their phone.
When you think of hacking browsers, you perhaps think of V8 heap exploitation, deep-dive fuzzing, crazy sandbox escapes, and so on. But what if I told you that you can still find vulnerabilities in major browsers that don’t require any technical knowledge?
Threat modeling an AI system is not simply a matter of adding prompt injection to an existing threat model. AI applications connect models to data, identities, external content, and tools, creating risks that can be easy to overlook.
AI writes a lot of your code now, and some of it is vulnerable. Meanwhile, frontier models have started finding vulnerabilities autonomously and at scale. One recent run surfaced 23,000 of them across 1,000 open source projects, most absent from any CVE database.
Back in 2019 I was studying, and even once worked on, quantum computers. Q-Day seemed a long way off then. I spoke then about how when Q-Day comes, not only would a lot of our encryption be obsolete but it would already be too late to do anything about it because those encrypted messages will have been, probably already had been, stolen.
What do you do when a web app encrypts its entire user database with a static 3DES key and exposes that key through blind SQL injection? You extract it, byte by byte, over thousands of timing probes. In 2026, you don't do it alone.
Best practices for securing web applications has more or less reached concensus. Combined with improved security in browsers many of the traditional attacks like cross-site scripting (XSS) and cross-site request forgery (CSRF) fail even before they reach your application.
Threat modeling an AI system is not simply a matter of adding prompt injection to an existing threat model. AI applications connect models to data, identities, external content, and tools, creating risks that can be easy to overlook.
Security decision makers are overwhelmed with AI-driven vulnerability discovery, leading to the perception of a relentless "vulnpocalypse" against which organizations have no defense. In this session, we will critically analyze this idea using vulnerability data, KEV information, and similar sources.
2026 set a new record with over 60,000 CVEs, a staggering 40% increase from the previous year. But raw vulnerability volume is only part of the story.