Agentic AI: Prof. Thomas Le Goff represents the FinAI Lab at BNP Paribas’ AI Summer Camp
01 September 2026
This year’s theme was Agentic AI.
How to Deploy Agents at Scale, Responsibly and Securely?
Télécom Paris’ Professor Thomas Le Goff, co-director of the FinAI-Lab (the joint laboratory between Télécom Paris and BNP Paribas), participated in the panel “How to Deploy Agents at Scale, Responsibly and Securely?” together with André Balleyguier (Applied AI leader at Anthropic) and Hanah-Marie Darley (CEO at UK’s start-up Geordie), moderated by Jérome Lebecq (Responsible AI Officer at BNP Paribas).
The panel highlighted some solutions and their limits. Anthropic presented its approach of AI internal governance through four layers of controls for agentic systems: model, harness (system design), environment, and humans.
One other key points discussed was human control and its limits: cognitive biases make systematic human validation not always the right way to mitigate risks. Humans should be strategically placed in the loop, e.g. for irreversible decisions taken by agents.
Finally, another key point of discussion was sustainability. Reasoning capabilities increase the energy needs by more than 100 times according to the most recent studies. Once again, we need to move away from the « bigger is better » paradigm towards a « smaller is sexier » one: the fact that you CAN use an agentic system for a task doesn’t mean it is the most efficient and appropriate for it.
This event was the opportunity to showcase the work conducted within the FinAI Lab, the joint laboratory between BNP Paribas and Télécom Paris, co-directed by Albert Bifet, Thomas Le Goff (Télécom Paris), Mariam Barry and Léa Déléris (BNP Paribas).
Established in 2025, the FinAI-Lab (Financial AI Laboratory) is a joint scientific research laboratory founded by BNP Paribas and Télécom Paris (a member of the Institut Polytechnique de Paris and the Institut Mines Télécom), dedicated to the study and development of artificial intelligence applied to banking and finance. It aims to address the scientific, technological, and regulatory challenges associated with the deployment of AI in critical and highly regulated financial environments.