KGeN unveils new platform 'KAI'…steps up 'Human Layer' strategy targeting the AI industry

Decentralized validation distributed protocol KGeN said on the 27th that it has unveiled a new initiative, 'KAI (Kratos AI)', aimed at the artificial intelligence (AI) industry.

KAI is a human-in-the-loop data and feedback platform that directly connects KGeN’s large network of verified users to AI training. Real users evaluate AI model outputs and provide feedback, helping improve accuracy and human-likeness through behavioral data and subjective judgments.

KGeN plans to leverage KAI as a 'Human Layer'—a core element required for AI training—by tapping its large community of verified users. The initiative focuses on improving model accuracy and human-likeness through reinforcement learning from human feedback (RLHF), subjective evaluations, and the generation of behavioral data.

KAI is designed around real human participation data, not synthetic data or automated traffic. KGeN verifies the identity of data contributors through on-chain reputation and validated user histories, a strategy it says will reduce so-called model-collapse risk, where AI models repeatedly train on AI-generated data and performance deteriorates.

In particular, KAI is structured to provide large-scale RLHF. Users deliver rapid evaluations and subjective feedback on model outputs, and the data is used to improve model accuracy and human-likeness in real time. KGeN’s vision is to elevate human judgment from an auxiliary input in AI training to core infrastructure.

KAI also highlights AI agent evaluation using gaming environments as a key use case. Users directly test AI agents in digital environments that require complex judgment and reasoning, and the resulting data is used as indicators to assess real-world responsiveness and reasoning performance.

"The future of AI depends not on more compute, but on better data," said KGeN founder Manish Agarwal, calling KAI "a platform that structurally supplies the human signals needed for AI training through participation from verified real users."

He added, "Through KAI, we will turn millions of users from passive consumers of the AI economy into active participants, and build a structure in which digital-native humans—not bots—train the next generation of AI."