Porcupine Labs is an independent nonprofit research organization based in the San Francisco Bay Area. Originating from Catalyze Impact, the lab addresses emerging biosecurity and operational safety challenges at the boundary where frontier AI systems direct physical wet-lab hardware.

As cloud laboratories, robotic liquid handlers, and automated synthesis platforms scale, digital model evaluations alone can no longer capture real-world risk. We work directly with platform operators, researchers, and policymakers to identify and close physical-layer vulnerabilities before deployment.

Our work centers on empirical, execution-layer evaluation across automated experimental workflows. Rather than evaluating biological risk solely through model weight benchmarks, we test how agentic planners, laboratory execution software, and robotic instrumentation behave under adversarial or ambiguous instructions.

We translate these empirical findings into concrete technical safeguards, including hardware-level telemetry verification, protocol anomaly detection, and auditable provenance standards for automated wet-lab infrastructure.

Our mission is to ensure that autonomous biological research infrastructure develops with verifiable physical-layer safeguards. By building rigorous evaluation frameworks and actionable defense strategies, we aim to enable safe scientific acceleration without expanding biological risk.