Michael Polansky is training an AI model on skin that’s still alive

Polansky’s startup sources discarded human skin from plastic‑surgery procedures through vetted biobanks such as the National Disease Research Interchange and the Cooperative Human Tissue Network, paying only cost‑recovery fees. After de‑identifying the specimens, the company’s proprietary culture system supplies nutrients and removes waste, stretching viability from the industry‑standard few days to a full month while preserving the tissue’s original cellular architecture, including epidermal, stromal, and immune signatures. This extended window lets researchers expose the skin to UV‑B radiation and monitor inflammation, collagen remodeling, and barrier repair over weeks—processes that previously could not be captured in vitro.

The move reflects a broader push to close the gap between software‑driven AI and wet‑lab biology. Traditional drug‑discovery pipelines rely on animal models or organoids that often fail to predict human responses, prompting firms like Insilico Medicine and DeepMind to seek richer human data. By maintaining ex‑vivo human tissue for longer periods, Outer Biosciences offers a middle ground: real human biology without the ethical and logistical hurdles of live‑patient trials. This capability could give the company a competitive edge in dermatology, cosmetics, and early‑stage toxicity testing, especially as AI models demand large, high‑quality datasets to improve predictive power.

Looking ahead, the platform’s success hinges on scaling tissue acquisition without compromising consent or IRB compliance, and on demonstrating that AI predictions derived from month‑old skin match outcomes in fresh tissue or clinical settings. Regulatory scrutiny may intensify around the use of human biospecimens for AI training, and any breach could jeopardize donor trust. Moreover, the claim that month‑old skin “looks a lot like day‑one tissue” will need rigorous validation to ensure that subtle molecular drift does not bias model outputs. Watch for partnerships with pharma or cosmetic brands that could provide real‑world benchmarks and for peer‑reviewed studies confirming the system’s fidelity.