AI x Biosecurity: Building Evaluations and Safeguards for High-Risk Biological AI Capabilities

In this talk, Peter Peneder will give an overview of his work at SecureBio on evaluating and mitigating biosecurity risks from advanced AI systems. This includes threat models for AI misuse, measuring AI uplift in biological workflows, developing new benchmarks for agentic and potentially superhuman bio-capabilities, studying LLM interfaces with biological tools, and exploring safeguards such as data controls and output filtering.
Date and Time: 25 June 2026, 4:00 pm (CET)
Location: Café Kreuzberg, 1070 Vienna
Language: English