DeepBrain AI launches AI Guard safety engine for generative AI
DeepBrain AI says AI Guard inspects inputs and outputs, with Face ID, malicious input filtering and deepfake detection via 10 APIs.
DeepBrain AI has completed development of its AI Guard safety engine and begun supplying it to companies and public institutions, according to venturesquare.net.
AI Guard inspects both the input information and the generated output in generative AI services. Its three core functions are Face ID identification, malicious input filtering and deepfake detection, provided through a total of 10 APIs that share the same authentication system and response format.
The engine applies a Fail-Closed approach, automatically blocking requests when model responses error out or exceed the processing time limit. DeepBrain AI also designed AI Guard to run its language and safety classification models on the customer's own GPUs, without calling external commercial AI APIs.
Testing and development
WiseStone, a KOLAS-accredited testing institution, tested AI Guard in October 2025 under the KS X ISO/IEC 25023 standard, where the engine achieved 96% accuracy in synthetic content detection under the test conditions.
DeepBrain AI developed AI Guard through a culture technology R&D project supported by the Ministry of Culture, Sports and Tourism and the Korea Creative Content Agency.
Quick answers
What is DeepBrain AI's AI Guard?
It is a safety engine that inspects input information and generated output in generative AI services, offering Face ID identification, malicious input filtering and deepfake detection.
How was AI Guard tested?
WiseStone, a KOLAS-accredited testing institution, tested it in October 2025 under the KS X ISO/IEC 25023 standard, and AI Guard achieved 96% accuracy in synthetic content detection under the test conditions.
Does AI Guard call external AI APIs?
No. It is designed to run its language and safety classification models on the customer's own GPUs without calling external commercial AI APIs.