Tynapse

Tynapse is a Seoul-based company developing runtime security and governance infrastructure for AI agents, which it markets as an "AI Trust Layer." The product sits between enterprise applications and any large language model, inspecting requests, decisions and tool calls before execution, applying policy checks, and blocking or intervening in real time. It is aimed at risks that the company argues surface-level traffic monitoring and prompt filters miss: jailbreaks and prompt injection, hallucinations and business misinformation, unauthorized or erroneous agent actions taken through tool use, data leakage and privilege escalation.

A second element of the offering is evidentiary: every decision the system makes is logged automatically into what the company calls a Trust Attestation Set, an audit record intended to demonstrate why an AI system acted as it did and to be submitted directly in audits. The company frames this against explainability and accountability obligations under regimes including the EU AI Act and South Korea's AI Basic Act, and states that its coverage is designed to align with AI governance, security and safety frameworks across North America, South America, Europe, the Middle East and Africa, Asia and Oceania — including NERC CIP, HIPAA and FDA AI/ML SaMD guidance, Brazil's PL 2338/2023 and LGPD, the EU NIS2 Directive, Saudi Arabia's NCA Essential Controls, Korea Financial Security Institute frameworks, Singapore's Model AI Governance Framework, Japan's METI AI Guidelines, and Australia's SOCI Act and Privacy Act reforms.

Tynapse markets to regulated and mission-critical sectors: financial services, public sector, healthcare, energy and infrastructure, manufacturing and mobility, and retail and subscription businesses. As of April 2026 it was running proof-of-concept trials with major Korean commercial banks and preparing for field deployment, with stated plans to expand from financial services into healthcare, the public sector and broader enterprise markets.

Founding story

The founding team is led by CEO Minseung Kang, previously a CTO in the financial sector with research and execution experience in AI trust and safety. The team also includes AI specialists from Google Research and engineers with backgrounds operating large-scale traffic systems; Kim Jineok is identified as CTO.

Business model

Tynapse sells enterprise infrastructure software deployed as an intermediary layer between customer applications and LLMs. Its stated engagement pattern is direct work with enterprise, financial and public-sector teams to deploy the Trust Layer, with the company describing deployment timelines measured in weeks.

Traction

Co-developing an AI agent runtime guardrail validation framework with KB Kookmin Bank, Korea's largest commercial bank, at proof-of-concept stage with a production path under evaluation; additional PoC trials reported with major Korean commercial banks. The company reached top finalist ranks — reported as the only Asian entrant — at the NVIDIA GTC 2026 AWS Startup Pitch and the 2026 Snowflake Startup Challenge, and exhibited at eGiSEC 2026, which drew a cumulative 26,000 visitors and participants from 19 countries.

Latest developments

Following the April 2026 seed announcement, the company said it would accelerate hiring of core talent and technology development. In July 2026 it presented multi-agent attack attribution research at an ICML 2026 workshop, and in August 2026 it published analysis of AI agent control failures, arguing that agent control breaks down both when high-level task approvals are stretched to cover newly generated high-risk actions and when decisions about those actions fail to prevent execution.