Zero-days are becoming industrial. What separates the US and China isn’t talent or even frontier AI algorithms. It’s structure: how each side can collect, route, and reuse zero-days. In a cyber arms race, the pipeline is the weapon. Read the brief to learn why. https://t.co/ud4nyIRsAl
🚨 Cato MDR Alert: We’ve detected a malicious IP address being used to download Quasar RAT, MeshAgent, and Defender Control via cURL.
Key IoCs, URLs and hashes in the thread below. 🧵👇
#ThreatIntel#QuasarRAT#MeshAgent#DefenderControl
Attackers don’t need bespoke malware for sophisticated intrusions—open-source frameworks are often enough. Cato CTRL blocked a suspected China-linked intrusion using TencShell, an undocumented Rshell-derived implant delivered through a masqueraded .woff payload and in-memory execution. If third-party access risk is on your radar, this report is worth a read: https://t.co/7N2mOkGOoY
🚨 New threat research: Cato CTRL found high-severity vulnerabilities in NVIDIA NeMo and Meta PyTorch that turn model files into RCE vectors. These models are pulled into systems with cloud credentials, IAM roles and sensitive data. Read more: https://t.co/4aRVNubiHZ
AI is moving fast, and the threat landscape is evolving just as quickly. Join Cato CTRL and Microsoft experts live on May 20 at SASEfy 2026 for a practical look at Agentic Security: where it started, where it is today, and what’s next. Real risks, real examples. Save your spot today 👉 https://t.co/OgC0UQ6kIe
With rising tensions in the Middle East, cyber activity from threat actors is expected to increase.
We are monitoring the threat landscape and has outlined a summary of notable Iranian-linked CVEs.
Read more: https://t.co/QXrPXPUlMe
Positive feedback can have odd side effects. A dog learned that saving kids from the Seine got him rewarded, so he started pulling them in to “save” them. AI models optimize the same way. Cato’s AI in Cybersecurity course shows how to manage the risks. https://t.co/mVSa5t7lL3
You secured the network. You secured the endpoints. Cato CTRL’s 2026 Threat Report reveals how threat actors now live inside your AI workflows- weaponizing them and walking out with your data. 5 discoveries that should keep you up at night https://t.co/sP780dvrzA
Internet-exposed MongoDB servers can be crashed. 🚨 Cato CTRL found a vuln putting 210,000+ instances at risk, letting attackers take databases offline in seconds, even in enterprise environments. A fix is available. Patch now. ⚠️ https://t.co/jyDV63HeYf
When your AI agent finishes a workflow, does it get offboarded?
We don't give employees unrestricted access. We don't leave accounts active after someone leaves.
But we do both with AI agents. That's where the risk begins. Time to treat AI agents as digital employees, not applications.
https://t.co/AzvC4GvjZ9
🚨 New threat research: Cato CTRL identified a threat actor selling root shell access to a UK automation company. The real prize wasn’t the root shell, but the CEO’s AI personal assistant based on OpenClaw (packed with sensitive data). https://t.co/yaIMbyI8H2