What question did this study set out to answer?

This study aims to introduce and validate the Open Agent Trust Stack (OATS) for secure AI agent execution.

May 21, 2026Open Access

Open Agent Trust Stack (OATS): A System Specification for Zero-Trust AI Agent Execution

View Full Paper

Ask AI

Bookmark

Authors

Randomized trial evaluates security measures for AI agents, suggesting a robust specification framework.

This study aims to introduce and validate the Open Agent Trust Stack (OATS) for secure AI agent execution.
Developed a five-layer architecture including the ORGA reasoning loop and cryptographic identity stack.
Empirical validation of core conformance requirements across nine hosted LLMs.
Assessment of forbidden tool-call attempts in cloud-adversarial environments.
263 forbidden tool-call attempts refused across 874 cloud-adversarial runs.
100% bite-rate against eight argument-injection attempts on hostile inputs.
0/560 escapes on pure-action vectors against Symbiont reference runtime.

Jascha Wanger (2026) studied this question.

View Full Paper

Ask AI

Bookmark

May 21, 2026Open Access

View Full Paper

Ask AI

Bookmark

Authors

Randomized trial evaluates security measures for AI agents, suggesting a robust specification framework.

This study aims to introduce and validate the Open Agent Trust Stack (OATS) for secure AI agent execution.
Developed a five-layer architecture including the ORGA reasoning loop and cryptographic identity stack.
Empirical validation of core conformance requirements across nine hosted LLMs.
Assessment of forbidden tool-call attempts in cloud-adversarial environments.
263 forbidden tool-call attempts refused across 874 cloud-adversarial runs.
100% bite-rate against eight argument-injection attempts on hostile inputs.
0/560 escapes on pure-action vectors against Symbiont reference runtime.

Jascha Wanger (2026) studied this question.

View Full Paper

Ask AI

Bookmark