- AngelsRound
- Posts
- Traceforce 🛡️
Traceforce 🛡️
AI agent security for enterprise devices
Spotlight
What if your security stack can't see the AI agents already running on employee laptops?
Quick Pitch: Traceforce gives security teams real time visibility and control over AI agents running on employee devices, blocking unsafe actions before they reach the network.
The Problem
AI on the Endpoint: AI agents can access files, databases, code, and credentials directly from employee devices.
Security Blind Spot: Network tools can't see agent activity happening locally on devices.
MCP Expands Risk: MCP gives AI agents access to more internal systems, expanding the attack surface.
Why It Matters
AI agents are gaining access to files, code, databases, and internal systems as enterprise adoption grows.
In the US, regulated industries including healthcare, financial services, and legal employ roughly 30 million people. Globally, the enterprise workforce spans an estimated 300 million employees, expanding the scope of AI governance and data protection.
Snapshot
Industry: AI Security
Headquarters: San Francisco, CA
Founded: 2025 (YC S25)
Funding: Raising $3M (Seed), previously raised $700K (Pre-Seed)
Backed By: Y Combinator, Antler
Traction: 3,000+ devices across six enterprises; 40+ active pilots
Target Customers: Enterprise security, IT, and AI platform teams
Founder’s Edge
Xia Hua, Co-Founder, CEO: Former Director of Engineering at Clumio, managing its ransomware protection products. PhD in Applied Math from MIT.
Varun Wadhwa, Co-Founder, CTO: Former engineer at LinkedIn, Clumio, and Microsoft. BS in EECS from UC Berkeley.
Playing Field
Network Security: Netskope and Zscaler operate primarily at the gateway layer.
Endpoint Security: CrowdStrike already has enterprise endpoint distribution.
AI Security: Emerging vendors secure AI through gateways, browsers, and APIs.
Traceforce's Edge: Built on device, where AI agents actually execute.
Analysis
Bulls Case 📈
Strong founder market fit from years building security infrastructure at Clumio.
The problem was validated with 50+ CISOs and CIOs before building.
On device architecture creates differentiation from gateway focused security tools.
Open source MCP tooling builds credibility around an emerging attack surface.
Bears Case 📉
Incumbents could extend endpoint products into AI agent security.
AI governance may become a feature, not a standalone category.
Established security vendors already own enterprise distribution.
Long term budgets for AI specific security remain unproven.
Verdict
AI agents are turning employee laptops from access points into execution environments.
Incumbents already own the endpoint, but they were built to detect traditional threats, not govern autonomous software. If agent behavior becomes a distinct security problem, the advantage may not be who is already installed, but who understands what the agent is doing.
The Startup Pulse
Another happening week in startup funding. Three signals from this week:
Inference is becoming a major market separate from training
AI infrastructure opportunities are spreading well beyond GPUs
Sovereignty is turning into real venture dollars across AI and space
Mistral: Raised €3B at a €21B+ valuation, led by Samsung. Europe is investing heavily in its own AI stack.
The Boring Company: Secured $3B at a $23B valuation to scale major tunnel projects across the UAE and U.S.Raised $3B at $23B. Elon Musk's tunnel startup is turning huge infrastructure projects in the UAE and U.S. into a venture scale bet.
Cognition: Landed $2B+ at a $48B valuation, led by a16z and Accel. Devin is nearing $900M in annualized revenue, keeping AI coding among the most valuable AI categories.
Strategic breakdowns of how the biggest AI companies actually work — their business models, their moats, their vulnerabilities. This week: Microsoft didn’t build AI. It didn’t need to.
Written by Ashher
