Perspectives: Multi-Agent Scientific Discovery Requires System-Level Safety

As AI agents automate more of the scientific pipeline, safety becomes a property of the agent collective rather than of any individual agent. Compositions of individually reasonable actions, taken by well-aligned agents, can still produce hazardous outcomes in settings like laboratories, biological design workflows, and cybersecurity. We argue that safety in AI-for-science therefore requires system-level design for multi-principal, multi-agent scientific ecosystems. To address this, we introduce SafeSearch, an abstraction casting scientific discovery as multi-agent search over a space where both value and hazard emerge compositionally, and use it to frame mechanism-design questions for safe scientific coordination.