Berry Blanton

I always wanted to be an astronaut. Like many kids, I looked at space and thought it was the future. Life took me in a different direction, and I ended up building a cybersecurity company. Recently I met Jeff Goldblum in Rome, and somehow my mind went back to Independence Day, malware, and space. At first it felt funny. Then I started thinking about it more seriously. With SpaceX, Starship, Starlink, and space exploration moving so fast, I believe cybersecurity in space needs a lot more attention. Satellites, spacecraft, ground stations, software updates, data transfers, AI, all of this creates risk. And in space, you cannot always patch later. You cannot always connect back. You cannot always rely on detection. So we decided to test something different. We sent thousands of malware samples to space to prove a point: malware prevention needs to work in the spacecraft, offline, under extreme conditions, including G forces, temperature changes, and barometric pressure. Maybe I should apply for a Guinness World Record for that. This video shows part of the journey, and the blog explains why I think this topic matters. https://lnkd.in/d2ZjhUqq #cybersecurity #space #SpaceX #AI #criticalinfrastructure #OPSWAT

The threat didn't break through the gate. It was welcomed in. For decades, the industry trusted files before anyone looked inside them. In Cybersecurity Upside Down, OPSWAT Founder and CEO .Benny .Czarny captures why that has to change. Available via audiobook, e-book, and print. https://bit.ly/3Mx4hwp

Vendors claim leadership. G2 customers confirm it. We're thrilled to earn Leader status in Managed File Transfer this spring, ranked on real reviews from real buyers in regulated industries. Learn more. https://bit.ly/3ODG602

Google DeepMind just published a paper on AI Agent Traps, and it highlights a structural issue that most of the industry is still underestimating. Link here https://lnkd.in/ezkfXDJW Paper: https://lnkd.in/eJb9ey2w You do not need to compromise the ai model to control its behavior, you only need to control the data it consumes, and that is exactly where the problem starts because the data flowing into AI systems today is not sanitized before it reaches the model. Agents are consuming websites, PDFs, emails, APIs, and other external sources, and they treat all of this input as trusted even though it may contain hidden instructions, embedded logic, or manipulated content that directly influences how the agent reasons and acts. This is not a model failure, it is an input integrity failure, because the agent has no awareness that it was manipulated and the user has no visibility into what the agent actually saw, which leads to silent compromise of decisions rather than something obvious that can be detected and stopped. We have seen this pattern before in cybersecurity, where years of investment in detection did not solve the problem because attackers consistently found ways around it, and the real shift only happened when we assumed every file could be malicious and rebuilt it into something safe before allowing it into the environment. That is why the technology Content Disarm and Reconstruction (CDR) exists Applied to AI systems, this means the control point is the intake layer, where data needs to be be sanitized before it reaches the data lake, the RAG pipeline, or the LLM itself, by stripping active content, removing scripts and embedded objects, and reconstructing files into deterministic safe formats that eliminate the ability to carry hidden instructions. At the same time, every external connector, whether it is web, email, or API driven, should be treated as untrusted and passed through the same sanitization process before entering prompts or memory. Antivirus and sandboxing still play a role, but they provide signals about what might be wrong, while CDR enforces that the content is safe, which is a fundamentally different outcome because detection can miss but sanitization removes the class of risk entirely. I wrote about this in 📘 Cybersecurity Upside Down https://lnkd.in/e5YUMdSj Not to repeat a message, but because this shift from detection to prevention is now becoming critical with AI. If your AI system is consuming unsanitized data, then by design you are allowing untrusted input to influence autonomous decisions, and that is the risk that this paper is clearly exposing. #CyberSecurity #AI #CDR #AgentSecurity #ZeroTrust #AIThreats

50 seats left and counting. Be among the first to experience Into the Breach: Breaking the Firewall featuring Kari Byron from MythBusters. Step outside the #RSA2026 show floor and into one of San Francisco’s most historic theaters for a cybersecurity experience like no other. You’ll discover that sometimes…the truth is scarier than fiction. https://bit.ly/4uqoftB