Join us in London or online on November 16 for the fourth annual QuantMinds-CompatibL Hackathon, focused this year on AI Safety and Security.
The event will take place at the InterContinental O2, London, from 09:00 to 17:00 GMT, with online participation available via Zoom.
Participants will compete to make AI agents more resilient to adversarial prompts across four streams, each focusing on an important application of AI for the buy- and sell-side, with no coding skills required.
Addressing AI Safety Through Prompt Engineering
The hackathon follows a two-team format. Each stream is an AI safety and security exercise where the red team plays the attacker, trying to get AI agents to provide incorrect responses, and the blue team plays the defender, working to stop these attacks.
Both teams will influence the outcome through prompts only. The organizers will be the red team. You will be the blue team.
Each stream will have its own task, dataset, and a set of red-team prompts. Solutions will be scored on the fraction of runs in which participants’ prompts restore the correct AI response despite the red-team prompt.
The competition will use the Claude Haiku 5.5 model. Participants are encouraged to obtain their own Anthropic API key in advance, as it will be required during the competition.
The hackathon is open to industry professionals, academics, and students, and both individual and team entries are welcome. Participants should bring their own laptop for the hackathon.
Awards and Registration
Award certificates will be presented to the top three places in each stream, plus a separate student award (individual or team, PhD level included).
The Grand Prize winner will receive a free pass to QuantMinds International 2027 (one pass per team).
Follow this link to learn more and register.
