01
AI security
threat-modeling · prompt-injection · data-poisoning · retrieval-acl · mcp-scopes
A language model gets access to confidential data or internal systems and has to stand up to an audit.
Securing RAG pipelines and LLM applications: who gets to see which documents, which inputs may steer the model and which actions an agent may take. With project experience up to the high-security zone of a bank.
- Threat modeling for LLM and agent applications
- Prompt injection and data poisoning in RAG pipelines
- Permission checks at retrieval time
- Limiting the tool and MCP access of agents
- Hardened runtimes for models and gateways