Language-Based Agent Control: A New Programming Model
A new paper introduces language-based agent control (LBAC), a programming model that applies techniques from programming languages and language-based security to agentic applications. LBAC requires agents to generate well-typed programs within the context of surrounding scaffolding code, allowing type-checkers to reject unsafe programs before execution. This extends guarantees such as access control, information flow, and data provenance uniformly across both agent-generated behavior and developer-written scaffolding, while preserving substantial expressiveness. The paper is available on arXiv under identifier 2605.12863.
Key facts
- LBAC is a new programming model for agentic applications.
- It brings techniques from programming languages and language-based security to agent control.
- Agents must generate programs that are well typed in the context of surrounding scaffolding code.
- Unsafe programs are rejected by the type-checker before execution.
- Policies apply uniformly across the entire application.
- LBAC preserves substantial expressiveness.
- The paper is on arXiv with ID 2605.12863.
- The announcement type is cross.
Entities
Institutions
- arXiv