AI security startup Tynapse said on Sept. 30 it was selected among the final top 10 at the Asia-Pacific finals of the Nvidia Inception Grand Challenge 2026, held in Singapore on Sept. 22. It achieved the result about a year after its founding in 2025.

The Nvidia Inception Grand Challenge is a competition where AI startups in the Asia-Pacific region showcase technologies and solutions that can be applied in real corporate environments. About 10,000 startups in the region participated this year and were evaluated on technological capabilities and business viability.

Founded in 2025, Tynapse is developing technology to verify and control the decisions and actions of enterprise AI agents. It has tested its technology in real corporate environments, including the financial sector, and introduced a technical architecture at the finals aimed at applying it to large-scale services.

Tynapse presented under the theme "AI Can Act. Tynapse Verifies If It Should". It said that even if an AI agent has the authority and capability to perform specific tasks, a separate verification is needed before it acts to determine whether the decision aligns with corporate policies and work standards.

The company said that as AI moves beyond answering questions to handling data and calling external tools to perform tasks directly, existing identity and access management alone makes it difficult to control AI decisions and actions.

To address this, Tynapse has developed what it calls a "decision trust infrastructure" that places a separate verification layer between an AI system's decision and execution. The approach verifies whether an action should be executed by checking the basis for the decision, corporate policies, work context and risk level just before the AI acts.

Based on the verification results, it either allows execution as PASS or blocks it as BLOCK, and it records the basis and policies used in the decision to enable post-audits. It said it focuses on verifying not only whether AI CAN act but also whether it SHOULD act.

It is also developing technology to process this in real time for large-scale enterprise services. It is building a runtime architecture to verify AI agents' decisions and determine whether to execute them by applying Nvidia Hopper-series graphics processing units, TensorRT-LLM and FP8-based inference optimisation.

Tynapse carried out a project with KB Kookmin Bank to verify the reliability of an AI financial advisory service. Their collaboration won the Financial Services Commission chairman's award, the "Cooperation Award of the Year", in June. It is currently expanding the scope of its technology applications from finance to areas including electronics and security.

Tynapse CEO Minseung Kang (강민승) said, "The fact that AI has the authority to access and act in a system, and whether the judgment it makes at that moment is correct, are completely different issues." He added, "Even if models keep changing, corporate policies and trust standards must be maintained."