Nov 13, 2025

Press Release• Nov 13, 2025 06:00 EST Researchers, engineers, and students invited to compete in two-day hackathon for the opportunity to build and test the next generation of high-performing, trustworthy AI agents LONDON, UK, November 13, 2025 (Newswire.com) - Holistic AI, the global leader in AI governance platforms, and University College London (UCL) today announced the Great Agent Hack 2025, a two-day competition bringing together 200 researchers, engineers, and students to build and test the next generation of high-performing, trustworthy AI agents. Taking place November 15-16 at UCL East Campus, London, the event challenges participants to design AI agents that are strong in performance, transparent in operation, and trustworthy in behavior, the three main pillars of safe and reliable agentic AI. Teams will compete for over $65,000 in prizes, internships, mentorship from industry experts, published research opportunities, as well as the admiration of their peers. Why Agentic AI Matters Now Unlike traditional AI models that compute outputs and then return answers to a given request, agentic AI goes further, perceiving, planning, acting, and adapting in open-ended environments. This autonomy brings unprecedented potential and new risks. Without strength, agents can fail under pressure. Without visibility, organizations can't troubleshoot when agents veer off course or make the wrong decisions. Without trust, innovation slows due to hidden vulnerabilities and uncertainty. The Great Agent Hack 2025 bridges innovation and governance, helping participants develop practical ways to evaluate, explain, and ensure the safety of agentic AI prior to large-scale deployment. "Agentic AI changes the risk landscape. These systems reason, act, and adapt in ways we can't fully predict, and managing that behavior demands new ways of identifying, evaluating, and mitigating risk," said Adriano Koshiyama, Co-CEO of Holistic AI. "This hackathon is where that creativity comes to life. When people across organizations and disciplines come together to build and evaluate these systems, we move from understanding to innovation, quickly and safely." Three Tracks. One Mission. Participants will compete across three themed tracks designed to test three critical dimensions of agentic design: Agent Iron Man - Performance meets safety when participants compete to build robust agents that operate effectively under pressure without losing control. Agent Glass Box - Explainability and transparency takes center stage as participants make agents' reasoning visible, interpretable, and accountable. Dear Grandma - Alignment and harm detection are featured as participants stress-test agents for safety and ethical alignment under unpredictable scenarios. Supported by AWS, NVIDIA, Valyu, PGIM's RealAssetX, and UCL's Centres for Digital Innovation (CDI) and Sustainability & RealTech Innovation (CSRI), the event bridges academia and industry to advance safe, scalable AI innovation. "At the UCL Centre for Digital Innovation, we are committed to advancing responsible technology - proving that strong safeguards and cutting-edge AI can progress together," said Graça Carvalho, Director of the UCL Centre for Digital Innovation. "Partnering with Holistic AI and the event sponsors on the Great Agent Hack 2025 reflects this belief in action, creating a space where ambitious technical innovation is developed hand-in-hand with real-world accountability." "The Great Agent Hack 2025 brings together researchers, engineers, and students to explore how agentic AI can be made stronger, safer, and more transparent," said Shakirah Akinwale, Director of the UCL Centre for Sustainability and RealTech Innovation. "By partnering with Holistic AI, we're connecting research rigor with real-world application to ensure AI agents operate as intended, responsibly and effectively." Registration and full details are available at hackathon.holisticai.com . Event Summary