The field of AI has a need for an established benchmark for developing and deploying federated foundation models, as well as a comprehensive understanding of the potential security risks involved. In particular, the challenge is to leverage a centralized foundation model like GPT-SW3, and investigate its potential evolution towards a federated model. 

This will require designing and demonstrating a next-generation infrastructure for training, testing, validating, and securely deploying federated foundation models. Alongside, assessing cybersecurity threats inherent to these distributed systems will be crucial. Ultimately, the goal is to contribute to the development of innovative solutions for the secure, federated deployment of AI models, setting the beginnings of new industry standards and guiding future advancements.

Two main goals:
 

• Create benchmark for training and, testing foundation models in a federated setup
   
• Impact Assessment of Adversarial Attacks on Federated Models and Possible defenses