Research
Our research teams investigate emerging threats, secure software, digital identity, and the societal impact of connected systems—so defenders can act on evidence, not fear.
Threat Research
Tracks adversaries, campaigns, exposure patterns, and the defensive signals organizations can use.
AI Security
Studies agent safety, model misuse, prompt attacks, and reliable controls for AI-enabled systems.
Privacy & Identity
Examines identity abuse, data exposure, verification, and privacy-preserving security systems.
Malware Analysis
Analyzes malicious files, infrastructure, and behavior to improve detection and response.
Vulnerability Research
Finds software weaknesses and develops practical, responsibly disclosed remediation guidance.
Prompt Injection Is an Architecture Problem, Not a Filter Problem
Every proposed fix for prompt injection that operates on the text of a prompt is defeated by rephrasing. The durable mitigations constrain what a model is permitted to do, not what it is allowed to read.
Publications
| Date | Research team | Title |
|---|---|---|
| Aug 15, 2026 | AI Security | Prompt Injection Is an Architecture Problem, Not a Filter Problem |