In the rapidly evolving landscape of artificial intelligence, many organizations express concerns about their readiness to implement generative AI responsibly. A recent report highlights that while a significant portion of businesses recognize the risks associated with AI, particularly in terms of explainability, only a small fraction are taking proactive steps to mitigate these challenges.
A startup based in Seoul has emerged as a key player in this arena, aiming to empower businesses to develop safer AI systems. Initially focused on data labeling, this innovative company is now expanding its offerings to include tools and data that facilitate the testing, monitoring, and enhancement of AI models without necessitating extensive technical knowledge. Recently, the startup successfully raised $15.5 million, bringing its total funding to around $28 million, with support from various investors.
The CEO of the startup, who previously worked as an AI researcher, was inspired to create a solution that simplifies the data labeling process. This led to the development of a unique app that incentivizes individuals to label data during their free time, allowing them to earn money while contributing to the AI ecosystem. The concept gained traction during a startup competition at a prestigious institute, where the team validated their idea and garnered initial interest.
Even before the app was fully operational, the startup managed to secure significant pre-contract sales, primarily from businesses associated with the institute. In its inaugural year, the company exceeded $1 million in revenue and established several important partnerships. Today, it boasts a diverse client base, including major corporations in South Korea, and has generated approximately $6 million in revenue in 2024.
As the company evolved, clients began requesting more advanced services beyond basic data labeling. This prompted the startup to pivot towards AI model evaluation, leading to the creation of Korea’s first benchmark dataset focused on AI trust and safety. The founders recognized the potential in this area and have since expanded their offerings to include pretraining datasets and evaluation services as the landscape of large language models continues to mature.
Recent developments in the industry, including significant investments in data-labeling companies, underscore the growing importance of this market. The competitive landscape is intensifying, with various players vying for dominance in AI training data solutions.
What sets this startup apart from its competitors is its unique approach to data sourcing and evaluation. By leveraging licensed datasets, particularly those derived from published literature, the company provides rich, structured insights that are often challenging to curate. Additionally, it offers a comprehensive evaluation platform that automates the generation of test data and assessments, ensuring that AI responses are safe and unbiased without requiring extensive manual intervention.
The startup’s ability to attract notable investors can be attributed to its strategic networking efforts, including hosting influential figures in the AI community. Following a successful event, the startup caught the attention of potential investors, leading to a fruitful funding process that spanned several months.
The newly acquired funding will be instrumental in accelerating research and development initiatives, particularly in the realm of automated evaluation tools for enterprise AI. Furthermore, the startup plans to expand its market presence in South Korea, Japan, and the United States, building on its existing workforce and establishing a foothold in Silicon Valley.