Discussion of OpenAI products and models from a security angle: misuse and threat reports, API and data handling, agent safety and integrating the tools securely.
🚨 Your AI application might be vulnerable even when its code is secure. As developers integrate Large Language Models (LLMs) into business applications, a new attack surface is emerging: prompt injection. Imagine building an AI assistant that can: - Read customer emails. - Sear…
I fail three out of four attempts at drawing a stick person, and the art on this post is 100% AI generated. I use fairly extensive prompting, and some images take several re-prompts to turn out the way I want. In general the process works for my needs. The art is one example of…
Victor Font — Keynote Speaker | Strategic Technology Leader | Fractional CIO | IT & Cyber Governance Expert | Published Author | FedRAMP Readiness | Board-Level Advisor | Digital Transformation | Enterprise WordPress/Formidable Forms
The AI-Native Enterprise Will Not Look Like Today’s Enterprise This is the twentieth and final installment of my AI Strategy Series, Building the Intelligent Enterprise. Throughout this series, I have argued that enterprise AI strategy is not fundamentally about selecting models…
Schomer — Principal Security Solutions Architect & Founder
When the email becomes the instruction An AI agent built to automate invoice processing was manipulated into changing vendor bank details - by a malicious email. No exploit chain. No traditional security alert fired. The agent read the email, treated its contents as work to do,…
Main Index - https://cyberr.ai/s/B3APl6f Flag: flag-073625 Vulnerability: Prompt Injection and Social Engineering Leading to Authorization Bypass The CTF was solved by manipulating the gatekeeper agent into accepting a user-asserted Supervisor Clearance ID without adequate verif…
What we know about the global Microsoft outage A major global Microsoft outage has caused problems, or in some cases disabled entire services. including emails and artificial intelligence products. Reports of problems with the email provider Outlook peaked globally at nearly 140…
Schomer — Principal Security Solutions Architect & Founder
I developed the ETBV AAV verification suite in a sandboxed Android environment. This forced me to optimize for resource constraints and test the framework's behavior in a mobile-first context, revealing edge cases in instruction validation that wouldn't appear in a standard Linu…
Cybersecurity threats: Europe is getting better. But so are attackers The EU has made significant progress in strengthening its cybersecurity. Still, persistent cyberattacks, geopolitical tensions and the rapid development of AI are creating increasingly complex risks, according…
UK to face more AI Cyber Attacks on Critical Infrastructure The United Kingdom is facing an increasingly complex cybersecurity threat, with hostile state-sponsored groups and sophisticated cybercriminals potentially targeting critical national infrastructure. The growing use of…
Do you think AI systems and AI agents are 100% secure? 🤔 In this video, we explore the latest cyber attacks targeting AI agents, such as indirect prompt injection, and how attackers exploit excessive permissions to compromise systems! 💡 Want to learn more about cybersecurity a…
AI Can Generate an Application Faster Than Ever. Can You Secure It Just as Quickly? AI coding assistants are changing software development. Developers can generate APIs, database models, authentication flows and entire application features with remarkable speed. But faster devel…
Are We Building Software Faster or Just Creating Technical Debt Faster? AI coding assistants have dramatically changed how quickly developers can produce code. A feature that previously required hours of implementation can now have a working first version in minutes. But here's…
Victor Font — Keynote Speaker | Strategic Technology Leader | Fractional CIO | IT & Cyber Governance Expert | Published Author | FedRAMP Readiness | Board-Level Advisor | Digital Transformation | Enterprise WordPress/Formidable Forms
What Should the CIO Own in the AI Enterprise? AI Strategy Series | Article 19 of 20 Building the Intelligent Enterprise As artificial intelligence becomes embedded in enterprise operations, the CIO's responsibilities are expanding. But there is a fundamental distinction that exe…
Google patched CVE-2026-125350, a use-after-free in Chrome's XML component that already has an exploit in the wild. That makes eight exploited Chrome zero-days this year. Update to 157.0.8103.50/.51 on Windows and macOS or 157.0.8103.50 on Linux, and watch for the Edge, Brave an…
CISA added Cisco BroadWorks CVE-2026-20320, an unauthenticated XXE fixed back in August, and Openwave Webtop CVE-2026-111793 to KEV. The BroadWorks bug was used in Operation Snowball, where an AI agent built on a jailbroken DeepSeek model hit a European telecom and pulled thousa…
Schomer — Principal Security Solutions Architect & Founder
Frontier LLMs aren't built to defend enterprise customers. They're built to burn compute. When you deploy a standard LLM inside an enterprise stack, the core hidden constraint isn't just accuracy—it's economic volatility. Chasing security postures through brute-force prompt engi…
An AI pipeline just reproduced 1,724 Linux kernel vulnerabilities for about $1.2M in compute. Your patch cycle was not designed for that. This week OpenAI described a run where an unreleased cyber model turned candidate kernel findings into reproduced bugs and working privilege…
🚀 Bridging the Gap Between Technical Controls and Corporate Governance 🚀 How does your organisation translate raw system configurations into executive-ready compliance data? Too often, GRC teams are buried under manual, outdated spreadsheets to track compliance against major s…
The first thing attackers spent after the Northflank breach wasn't your database. It was your AI API keys. On 30 October the cloud platform told customers to assume every environment variable had been decrypted and to rotate all keys. Users soon reported their LLM keys being use…
Binance is warning iPhone users about worse_query, an iOS sandbox escape published on GitHub that lets an app read other apps' files on iOS 26 and 27, including wallet seed phrases, private keys and logins. Apple's updates from yesterday (iOS 27.2.1 and 26.7.3) don't list a fix,…
Jason Murray — Strategic Information Security Leader Championing Quantified Risk Analysis and Security Process Improvement.
Another step outside the echo chamber. Venkatesh Rao offers an analogy for AI as a new kind of cell: model weights as a genome compressed from the internet, tools as proteins, humans as mitochondria supplying "liveness" (stakes, valuation, knowing which outcomes matter). Ethan M…
Sam Curry and Shubham Shah could take over any customer account at two car brands through a ForgeRock OpenIDM admin console still set to admin:admin. A URL parsing trick got them past the edge block, and from there they unlocked and started a test car. Both were fixed about two…
Jason Murray — Strategic Information Security Leader Championing Quantified Risk Analysis and Security Process Improvement.
There is a belief that RSI -> AGI -> Doom! Yuval Levin asks whether we're even on the path to AGI: AI progress is getting narrower, not more general, and recursive self-improvement would likely deepen that. "The assumption that RSI would independently facilitate the emergence of…
I've been working on a personal project called Foxscan 🦊, and I wanted to share it with the cybersecurity community and get some feedback. The idea came from a problem I kept running into: different news outlets covering the same cybersecurity incident, each reporting different…