Security research and technical analysis
LLMSecure for GitHub Actions protects AI workflows from prompt injection in issues, PRs, and comments. A drop-in step that validates every input before your AI agent reads it.
Read articleWhat does it take to compromise an AI coding tool with millions of users? A GitHub issue, a clever title, and an AI agent that trusts what it reads.
Read articleWhat happens when an attacker scrambles a malicious prompt beyond recognition? Can you really trust your first line of defense? Intent extraction is a new approach that gives prompt filtering the depth it has been missing.
Read articleWhat if we treated prompts the way we treat executables? The idea behind llmsecure is simple: apply the principles of malware analysis to LLM security. Here is why I built it and how it works.
Read article