Producing a series of concise, one-page explainers on foundational AI safety and security concepts, tailored for stakeholders in French institutions.
About the project
General-Purpose AI (GPAI) systems are increasingly deployed across the economy, with frontier models showing rapid progress in reasoning, autonomy, and task generality. As capabilities grow, so do structural security and control challenges that remain poorly understood outside specialized technical communities. Policymakers often lack concise resources, especially in French, explaining the core mechanisms that can drive systemic AI risks.
This project proposes producing a series of one-page briefs that introduce the main concepts of AI security in a clear, policy-oriented format. All documents will be written in French.
The series would cover classic topics in the AI Safety and Security literature such as the limits of current interpretability techniques, situational awareness and evaluation awareness, reward hacking, model agency, autonomous replication, etc.
Theory of change
This project advances AI safety by helping us increase the conceptual readiness of French institutions that could push for international coordination on frontier AI oversight. We already have strong demand from many actors for our policy notes, but limited capacity to produce them. The faster we can write these briefs, the faster these institutions will mature on the topic.
Your role
Mentees will help research, draft, and refine the one-page briefs. They will work under our guidance (to identify relevant sources, and iterate on early drafts) but will have autonomy in structuring each brief. They may also contribute to selecting new topics as the series expands.
Prerequisites
- Strong understanding of AI risks and model behavior.
- Experience in scientific communication, ability to synthesize complex technical topics.
- Fluency in French.
Location preference
We have a clear preference for contributors who can come to our Paris office at least a few times, but this is open to discussion.
Application question(s)
- Create a diagram (hand-drawn, digital, or using any tool) that elegantly explains a core AI security concept of your choice (e.g., goal misgeneralization). Include a caption (in French). Your diagram should be understandable to a policymaker with limited technical background.
- Optional: Provide a link to one or more relevant writing samples, ideally demonstrating your ability to communicate technical material.
About the mentor
Jérémy Andréoletti is Head of Research at the General-Purpose AI Policy Lab, where he works on risk modeling, forecasting, and governance for advanced AI systems. He completed a PhD at ENS-PSL on macroevolutionary Bayesian modeling and co-founded EffiSciences to support students and researchers interested in high-impact work (AI Safety, biosecurity).