
100. Why Your Employees' Favorite AI Tool Might Be Leaking Your Data
About this episode
Every employee at your company probably has ChatGPT, Claude, and Gemini installed, and nobody's tracking what data goes where. Xia Hua, co-founder and CEO of Traceforce, came back a year after her first appearance to show us what that looks like from the inside. Her team's open source scanner, MCP X-Ray, found a prompt injection flaw in Playwright, one of the most widely used MCPs, and she triggered it live with a single sentence. We also get into Anthropic's report on the espionage campaign that used Claude and a set of MCPs against about 30 organizations. And the bigger problem underneath it all, that data and instructions are now co-mingled, so any tool that reads text can be told what to do by that text.
Xia: www.linkedin.com/in/xia-hua-ph-d TraceForce: www.traceforce.ai MCP X-Ray: www.github.com/traceforce/mcp-xray Jon: www.linkedin.com/in/jon-mclachlan Sasha: www.linkedin.com/in/aliaksandr-sinkevich YSecurity: www.ysecurity.io
🔒 Sponsored by YSecurity
Closing a big deal and getting hit with a vendor questionnaire, SOC 2 request, or pen test requirement? That's exactly what we handle.
YSecurity embeds a team of security engineers with backgrounds from Apple, Uber, Microsoft, Robinhood, and Brex directly into your company. No full-time hire, no bloated retainer. You set a monthly cap and we handle everything.
Book a free strategy call and we'll tell you exactly where you stand.
👉 Book your free call: 30 Min Meeting | YSecurity.io | Cal.com
👉 Learn more: YSecurity | On-Demand Cybersecurity Team for Startups — SOC 2 in 5 Months
Get every episode summarized
Each time The Security Podcast of Silicon Valley publishes, we email you a written briefing from the transcript — the topics, who appeared, and any specific claims, with the ad reads skipped.
Email me new episodesFree for 3 shows. No card needed.
Hosts & guests
No transcript yet
This episode has not been transcribed. Request it and it moves to the front of the queue.
More episodes
More from The Security Podcast of Silicon Valley

102. CISO does not spell CEO (with Chris Kirschke)
The Security Podcast of Silicon Valley

101. Hacking AI is now a TikTok skill (with Johnny Hung & Munam Wasi)
The Security Podcast of Silicon Valley

99. AI Onboarding Is a Security Problem, Not a Feature (with Ged Ossman)
The Security Podcast of Silicon Valley

98. How Browser Security Became the New Battleground for Enterprise AI
The Security Podcast of Silicon Valley