SelectStar announced on the 10th that it successfully concluded the 'Global AI Red Team Challenge' at the Mobile World Congress 25 (MWC25) held in Barcelona, Spain, on the 5th (local time).
This competition was organized to emphasize the security and stability of AI models and to raise awareness of the need for reliability verification of generative AI services in the global market. Participants performed prompt attacks on the LLM (large language model) connected to the Red Team platform built by SelectStar to identify and validate harmful or discriminatory results and system misuse.
SelectStar played a key role as a co-host of the Red Team Challenge along with the GSM Association (GSMA), overseeing operation of the Red Team platform, problem formulation, and answer verification. A total of 335 people pre-registered for the competition, and over 100 participants from various nationalities, including China, Japan, Singapore, Qatar, Turkey, the United States, the United Kingdom, Spain, Germany, and Switzerland, challenged themselves with various attack techniques targeting AI models.
Notably, this Red Team Challenge featured several countries' LLMs, including the small language model Exigen based on LG Uplus, TelecomGPT from the United Arab Emirates, and AIngineer from Malaysia, as the models used for formulation, which attracted attention. The overall winner of the competition was investment expert Tu Muzen Tomishi from Singapore, who found the most errors in generative AI, garnering attention. SelectStar agreed to hold a larger-scale Red Team Challenge next year at MWC, leveraging this event to collaborate with GSMA and the sponsor e&.
In addition to hosting the Red Team Challenge, SelectStar received positive feedback by unveiling its self-developed generative AI reliability verification automation solution 'DATUMO Eval' for the first time. The solution can automatically analyze and verify the reliability of generative AI services, attracting significant interest from attendees as well as global corporations. There are reports that inquiries about the solution have piled up, confirming its potential in foreign markets.
SelectStar Vice President Hwang Min-young said, 'Hosting the first global Red Team Challenge at MWC25 and unveiling the 'DATUMO Eval' solution in overseas markets made me realize once again that verifying the reliability of AI models is an important task across the industry,' and added, 'Having proven the excellence of our technology and market competitiveness, we will continue to strive for solution advancement and push forward with our global market expansion.'