SelectStar announced that it concluded the 'Global AI Red Team Challenge', a side event of 'MWC 2026' held in Barcelona, Spain at 11:15 AM (local time) on the 3rd.
This event, co-hosted with the GSMA for the second consecutive year, took place at the "Talent Arena" within the MWC venue. Approximately 130 participants, including AI experts, developers, and industry professionals from around the world, attended the event.
Participants attempted various prompt attacks to explore vulnerabilities in Large Language Models (LLMs). Seventy-eight participants submitted final results, marking a higher participation rate than last year. First place went to data engineer Haktan Kocyigit, who received €1,000 and a ticket to MWC 2027. Second and third place winners also received cash prizes and tickets to future events.
This year's challenge tested key LLMs developed by global telecommunications and AI companies. These models included SK Telecom's "AX K1," LG Uplus's "ixi-GEN," the United Arab Emirates' "TelecomGPT," AT&T's "Ask AT&T," and NetoAI's "Tslam."
This event was conducted in a non-competitive and confidential manner, with no public disclosure of rankings. Participants and judges were provided with de-identified company and model information. Rather than comparing models or disclosing vulnerabilities, the results were designed to be used by each company to improve its services.
The verification process utilized SelectStar's AI reliability assessment platform, Datumo. This platform supported stable operation and analysis of results even in environments where multiple users were simultaneously accessing the platform and conducting attacks in various languages and methods.
GSMA's Director of AI Technology, Lewis Powell, explained that this event was an opportunity to examine security vulnerabilities in generative AI and explore ways to improve them. SelectStar Vice President Hwang Min-young noted that demand for AI reliability verification is growing in industrial settings and announced that the company will continue to enhance its evaluation system to ensure it can be applied to real-world service environments.
Meanwhile, with the recent spread of generative AI, regulatory agencies and companies around the world are accelerating the establishment of model reliability and safety verification systems, and the related evaluation market is also growing rapidly.
- See more related articles
You must be logged in to post a comment.