Netskope

What happens after Google delists a malicious extension? 🔓 Sometimes it comes back. A Chrome extension removed in January 2026 for exfiltrating AI conversations was relisted, rebuilt its reputation with a clean update, then shipped a new build that opens an affiliate link on every update and every uninstall. Marketplace removal is not proof a threat is resolved, especially when the same extension ID can quietly return with a new payload. #ThreatLabs #ChromeExtension

"Context is the king."👑 That is how Mohit k, Sr. Application Security Engineer at Netskope, sums up what he learned red teaming Netskope's own client software with frontier AI models. Season 4 of Security Visionaries kicks off with Mohit joining host Emily Wearmouth to explain what it took to hunt for memory corruption bugs, some of the hardest and most dangerous flaws to catch, using AI models as the researcher. We covered: 🔹Why proving a bug against a real, running system matters more than trusting a model's word for it 🔹The isolated, offline sandbox labs built so AI agents could hunt for crashes safely 🔹Why phase two brought in a second, independent instance of Claude Mythos, given a completely fresh context window, to double-check every finding Links to the full episode are in the comments.

🤖 "For data's sake, give me guardrails." AI agents move fast. Your data security should too. See how Netskope gives AI the guardrails it needs to innovate at speed, without putting your data at risk.

Is your AI agent quietly turning your company into a regulated model provider? Under the EU AI Act, modifying a model and attaching an agent to it can shift you from deployer to provider, pulling in a new set of compliance articles. And Article 50, the baseline transparency requirement for agents interacting with people, was already due for compliance as of August 2, 2026. Our latest blog breaks down: 🔹Whether agents legally count as AI systems 🔹When a modified model changes your compliance category 🔹How Chapter III's high-risk rules get triggered Get the full picture, including a flowchart for your specific situation. Link is in the comments below. ⬇️

A network outage is not an option for Bechtel. Nor is high latency. The global engineering and construction leader serves customers on all seven continents, where engineering, procurement, and construction (EPC) tools must move data reliably between 3D modeling platforms and job sites, all while meeting FedRAMP requirements for its government business unit. With Netskope One Security Service Edge (SSE) and Netskope One AI Guardrails, Bechtel now has a single approach to managing traffic and enforcing security policy across the organization, mitigates prompt injection and jailbreak attempts in its AI tools, and gets strong performance worldwide through the Netskope NewEdge Network's 120+ data centers. “A change made in one place with Netskope applies everywhere. What we allow or block is consistent for all users. That’s been a tremendous improvement.” - Patrick Noon, Chief Information and Digital Officer, Bechtel

What do 28 malicious npm package versions across four unrelated namespaces have in common? 🔹An identical or functionally equivalent stealer payload 🔹A shared trick for resolving the exfiltration domain through an Ethereum smart contract 🔹A credential sweep built to hit developer machines and CI/CD build agents alike Static indicators alone won't catch this one. Detection has to fire on behavior. Read the full analysis from Netskope Threat Labs (link in comments ⬇️ )

It's time for the #WeekendWrapUp of Netskope news you may have missed: 🔸CRN | Cybersecurity Agent Collisions Are Coming And Could Be ‘Very Ugly’: Netskope CISO: https://lnkd.in/eQeUipGd 🔸Fox Business Network | Netskope CEO: We should have an AI kill switch, but it isn't the answer: https://lnkd.in/dKHEiry3 🔸MeriTalk | The First Federal Guidance on AI Agents Is Here. Most Agencies Can’t Meet It Yet.: https://lnkd.in/eWM7GD6Z 🔸Security Boulevard | Netskope Unveils DataSec Command Center for AI and Cloud Data Controls: https://lnkd.in/enH2TzM9 More news you may have missed ⏬ https://lnkd.in/gW5if9Nf

What happens at Black Hat doesn't have to stay in Vegas. Thank you to everyone who stopped by to "get the scoop" on what's new from Netskope. A few highlights from a busy show: 🔸 We launched Netskope One DataSec Command Center, unifying data security from discovery to remediation. 🔸 We joined NVIDIA's Open Secure AI Alliance as an inaugural partner. 🔸 We were named a Leader in both the 2026 Gartner® Magic Quadrant™ for SASE Platforms and the 2026 Gartner® Magic Quadrant™ for Security Service Edge. 🔸 We're also the only vendor among the three highest-scoring vendors for all use cases in the 2026 Gartner® Critical Capabilities for SASE Platforms report. Missed us in Vegas? We'd still love to connect: https://lnkd.in/gi4jCW7h

Our latest AI Index data reveals global adoption climbed 1.9% week over week, while meeting app usage declined. See the full dashboard for more insights: https://lnkd.in/gtpQi8A4 #AIIndex

"You need to zero trust every AI agent. It's your co-worker. But it's a co-worker with no conscience, with no EQ, with no morals. You as a company need to have controls within your own environment." In his appearance on 'The Claman Countdown' on Fox Business Network, Netskope CEO Sanjay Beri explained why an AI kill switch isn't the answer to stopping AI agents from going rogue. Watch his full take on why companies need their own defenses in place, with zero trust applied to every AI agent operating in their environment. https://lnkd.in/dKHEiry3