Tynapse Named to Top 10 in NVIDIA Inception Grand Challenge 2026, Puts Forward Pre-Action AI Decision Verification
TECHWORLD ·
✦ AI Summary
Tynapse was named to the Top10 of the Asia-Pacific finals for the NVIDIA Inception Grand Challenge 2026 about 1 year after its founding.
The competition is for Asia-Pacific AI startups and was held in Singapore on the 22nd as a venue for presenting AI technologies and solutions that can be applied in real corporate environments.
Tynapse proposed a structure that places a verification layer between AI judgment and actual execution, and is currently carrying out a project to verify the reliability of financial consulting AI with KB Kookmin Bank.
Tynapse was named to the final Top10 in the Asia-Pacific finals of the NVIDIA Inception Grand Challenge 2026 about 1 year after its founding. The selection was announced on the 30th.
The NVIDIA Inception Grand Challenge 2026 is a competition for Asia-Pacific AI startups. Held in Singapore on the 22nd, it is a venue for presenting AI technologies and solutions that can be applied in real corporate environments.
Tynapse is an AI security startup that verifies and controls the decision-making of enterprise AI agents in real time. Since its early days, Tynapse has applied AI judgment, action verification, and control technologies to enterprise environments, including finance.
Tynapse was selected for the final Top10 after a comprehensive evaluation of technology and business viability among about 10,000 participating startups. In the finals, Tynapse presented its experience in technology verification accumulated in real corporate environments and its technical architecture for large-scale service deployment.
For its final presentation, Tynapse presented the theme, "AI Can Act. Tynapse Verifies If It Should." Tynapse stressed that even if AI has the authority and capability to perform tasks, additional verification is needed before action to determine whether it complies with corporate policies and work standards. It also explained that, as AI advances, existing identity and access management have limitations, making sufficient control difficult.
As a solution, Tynapse proposed a structure that places a separate verification layer between AI judgment and actual execution. Through its independently developed Decision Trust Infrastructure, Tynapse said it checks the grounds for judgment, corporate policies, work context, and risk immediately before AI action, treating "whether this action should be allowed" as the subject of verification. Based on the verification result, execution is either passed (PASS) or blocked (BLOCK), and the basis for judgment and applied policies are recorded to make post-event audit and verification possible.
Tynapse focused on AI governance and proposed a direction that expands from verifying whether AI can act (CAN) to verifying whether it should act (SHOULD). It said this approach goes beyond determining whether AI is capable of doing something and checks whether it should actually be carried out.
Along with that, Tynapse presented as part of its final pitch a technical architecture for performing real-time decision verification for large-scale enterprise services. The company has applied NVIDIA Hopper GPU series, TensorRT-LLM, and FP8-based inference optimization. It is also building a runtime architecture to rapidly verify AI agent judgments and determine whether to execute them.
The technology is being verified in real corporate environments, centered on domestic financial institutions. It is currently carrying out a project with KB Kookmin Bank to verify the reliability of financial consulting AI. The Tynapse-KB Kookmin Bank collaboration won the Financial Services Commission chair's "Partnership Award of the Year" last June. Its scope of application is expanding beyond finance to industries such as electronics and security, where AI performs real work.
The company said that, based on the momentum of its Top10 selection, it has technology verified in the financial sector and plans to expand it into the global enterprise market.
In addition, Tynapse CEO Kang Min-seung said that AI having system access and action permissions is separate from the legitimacy of determining when to use them. He also pointed to the sustainability of model changes and stressed the need to maintain corporate policies and trust standards.
Source: TECHWORLD · Kim Hye-jin
Original: https://www.epnc.co.kr/news/articleView.html?idxno=407543
References
This article was produced with the help of an automated content generation algorithm.
Source: TECHWORLD
View originalThis article was summarized and organized by BizCrush based on the original article from TECHWORLD. For exact quotations and full details, please refer to the original article.