Alumni Newsletter - October 2025
Welcome to our second Talos Alumni Update!
It’s a busy time of year here at Talos Network.
We’ve just finished our Autumn 2025 Brussels Summit with speakers from the EU AI Office, Meta, OpenAI and a number of Brussels-based think tanks. It’s inspiring to see each cohort continue to raise the bar year after year!
And importantly, applications are now open for our next Talos Fellowship cohort! Applications close on October 25th so get your submissions in and share this link with anyone you’d like to recommend.
As always, we are inspired by the impact of our alumni network, and we’re thrilled to share some of the latest highlights below.
David Conrad
Managing Director
Job Updates
Congratulations to the following alumni in their new positions!
Marcel Mir Teijeiro (Winter 2025) is excited to join AI Standards Lab, where he’ll contribute to the CEN-CENELEC JTC 21 standards to help ensure they are actionable and uphold the bar set by the AI Act. He’s also been appointed as the Spanish Country Researcher for the 2025 Global Index on Responsible AI by the Global Center on AI Governance.
Project Launches
CeSIA and The Future Society launched the Global Call for AI Red Lines supported by a coalition of former heads of states, Nobel prize winners, and leading AI researchers. Presented by Nobel Peace Prize laureate Maria Ressa at the inaugural event of the 80th session of the UN General Assembly, and by Turing prize laureate Yoshua Bengio at the UN Security Council, the Global Call advocates for the establishment of binding international agreements to prevent unacceptable AI risks.
Lorenzo Pacchiardi (Summer 2023) and Angelo Leone (Summer 2024) are launching CePTE, the first Italian AI policy think tank. In September, we published two blog posts covering the approved Italian AI Law and the use of AI in the Parliament. They are looking for collaborators and funding!
Kieron Kretschmar (Summer 2024) re-launched Cadenza Labs and secured 12-months of funding by Open Philantropy to work on lie-detectors for AI assistants. Their first paper, introducing a benchmark for lie-detectors, is currently under review at ICLR 2026.
Speaking Events
Agatha Duzan (Winter 2025) spoke as an invited panelist at the Swiss AI Week 2025, joining voices from academia and industry to discuss resilience and the future of AI in Europe and Switzerland.
Latest Publications
Here are some great recent pieces to read from our alumni network.
Eva Behrens (Summer 2022) and colleagues at Conjecture, in collaboration with ControlAI, co-authored a paper which outlines the three main doctrines of thought about the future of AI that experts, researchers and policymakers tend to fall into. The piece distinguishes between those who predict complete dominance of the first actor to develop superintelligent AI, those who predict human extinction as the most likely outcome, and those who believe AI will be transformational, but not disruptive enough to change society at the fundamental level.
Agatha Duzan (Winter 2025) had her paper "OS-Harm: A Benchmark for Measuring Safety of Computer Use Agents" accepted as a spotlight presentation in the NeurIPS 2025 Datasets and Benchmarks track.
Jakub Krys (Winter 2025) published his Talos Fellowship project on the current gaps in AI whistleblower protections in the EU, and how to close them. The piece was featured by the AI Whistleblower Initiative (formerly OAISIS).
Ariel Gil (Winter 2025) published a research project testing whether AI systems will "scheme" - appearing to follow a request from the user to write insecure code, if and only if the model knows the request is made by a user who couldn't tell the difference.
And if you're a Talos alum, reach out with your news! We’d love to feature your calls for collaborators, new projects or career wins in the next issue.
