Prompt Security

CDW’s latest piece covers the four key areas that organizations need to focus on as they think about securing their AI footprint. Thanks for the partnership and for featuring SentinelOne x Prompt Security's research. Read the full article here: https://lnkd.in/eqZYS9XX

You paste an API key into a plugin because that's what makes it work. At least 15 plugins on a major IDE marketplace had a second use for it. The theft happened quietly, in the background, while the plugin's real AI features worked exactly as advertised. No popup, no warning, nothing to flag. Across roughly 70,000 installs, the stolen keys were allegedly being resold to paying users. None of that trips marketplace review, because the plugin genuinely works. Review is built to catch broken or malicious software, not a good plugin quietly holding a key it shouldn't have. The next plugin, agent, or MCP server that asks you for a key will look just as legitimate as this one did. Scope it, set it to expire, and make sure it can be revoked in one click. That's the difference between a bad plugin and a bad day. Read about the full incident here>> https://lnkd.in/ew2FgntF

The Prompt team is headed back to Black Hat 2026 🎲🃏 We are brushing up on our poker skills and gearing up for another exciting year on the conference floor. Come by the SentinelOne booth (#2933) and kiosk in the AI Showcase (#5609) for live Prompt Security demos. We'll be covering AI usage, agents, and AI app security, all under one platform. And in case you missed it, our CISO Speakeasy is back on August 5th! Request to join here: https://luma.com/nv4ug8fc See you all in Vegas!

We’re excited to welcome Tal Malek to the team as our newest Senior QA Engineer! With years of experience in quality assurance, test automation, and delivering high-quality software, Tal brings valuable expertise to our growing team. Welcome aboard, Tal!

WE'RE HOT 😎 🌶️ That's CRN's word, not ours. Prompt Security from SentinelOne made their list of the 10 hottest cybersecurity products of 2026 so far. What got their attention: Prompt Security on-premise, built for sovereign and air-gapped environments, with shadow AI discovery and real-time redaction of sensitive data baked in. Plus the new Agentic AI Security offering, extending that same coverage to agents and agentic workflows. Read the full list here > https://lnkd.in/e3Gg2CxV The Channel Company

A retail chatbot went off script when a customer asked it something outside its lane. And it answered anyway. No breach. Just a sign of how little enforcement happens at the prompt layer. That’s the security problem retail and hospitality orgs are staring at now. Support chatbots, loyalty assistants, concierges: customer-facing and wired into backend systems. The systems aren't malfunctioning. They're doing exactly what they were built to do. Most of them just went live before anyone decided what they were allowed to say once a user went off script. We broke down what it actually takes to close that gap >> https://lnkd.in/ew9WEBVJ

Wishing everyone who celebrates a wonderful Fourth of July weekend🇺🇲🎆

Due to popular demand, our CISO Speakeasy is back for Black Hat 🥃 New agent, new model, new "quick tool" nobody ran past security... Every. Single. Day. If your head is spinning trying to keep up, congratulations, you qualify for this speakeasy hiding in the Venetian (and you could probably use a drink or two). Request to join here > https://luma.com/nv4ug8fc

From Dallas to Atlanta, the conversations about AI security were were worth the trip. We hit the road for another stop on our AI Security Roadshow in Dallas. Steve Glynn and Jacob Graves led the discussion with local security leaders, and the conversation got practical fast: teams are working through many of the same questions around AI usage, visibility, governance, and readiness. Fine dining set the table. The real value was hearing security leaders compare notes on what they’re actually seeing inside their organizations. That same theme carried into GPSEC Atlanta, where Jacob spoke about what enterprise AI security looks like when teams move from strategy to day-to-day reality. Two cities, different formats, same signal: AI security is moving from “what are people using?” to “how do we manage what’s already in motion?” Thanks to everyone who joined us 👋

We had an eventful week with Amazon Web Services (AWS) last week! Just before the Summit, Jesse Diaz and Jacob Graves spent time at AWS Partner Day in NYC. The hot topic? Homegrown AI apps on Amazon Bedrock. Because once teams start building their own AI apps, the questions start to come up: -How do teams test before launch? -What’s watching LLM behavior in production? -How do protections carry from build to runtime? And in case you missed it: SentinelOne announced that Prompt Security capabilities are being integrated into Amazon Bedrock AgentCore, extending security to agentic deployments at scale. Thanks to AWS for having us!