Palace Cybersecurity secures large enterprises from employees abusing their access privileges in SaaS systems like Google Workspace or Microsoft SharePoint.

Like how YouTube or TikTok can recognize a video that you should watch but have not, the Palace Foundation model can recognize a corporate resource that an employee should not access, but did — having learned from past access patterns. The model assigns each user and resource an embedding vector. The two vectors point in the same direction if and only if the user would access the resource.

Applied to past accesses, the model highlights access anomalies, achieving state-of-the-art performance for insider threat detection and investigation. Applied to potential accesses, it powers embedding-based access control (EBAC), a learned access policy that complements RBAC by removing entitlements that hand-crafted configuration inevitably leaves open, either ahead of time (posture management) or in real time (dynamic access control).

The founder researched and deployed a similar system at Google. The corresponding research paper (FACADE) was accepted for publication at USENIX Security 2026 and was presented at Blackhat 2025. Palace is currently building the MVP and working with early design partners.

Palace Cybersecurity

Org chart

Teams

Offices

Palace Cybersecurity

Org chart

Teams

Offices

Related companies