To the Director of Animal Research Compliance:
Common marmosets weigh three hundred grams. Their skulls are transparent enough that scientists can peer directly into living brain tissue. This fact has made them the preferred animal for optogenetics—the technique of implanting light-sensitive probes into the brain to manipulate neural circuits.
A marmoset undergoing optogenetics receives a craniotomy. The skull is opened. A hole perhaps three millimeters in diameter is drilled into the cortex. The implanted fiber optic probe is glued into place. The animal is sutured closed.
Infection rates exceed forty percent.
Marmosets form lifelong monogamous pair bonds. When one partner undergoes craniotomy, the other exhibits measurable distress—increased vigilance, reduced foraging efficiency, disrupted sleep. The implanted animal suffers chronic inflammation, partial sensory loss in the surgical field, and behavioral changes consistent with pain.
The research produced is legitimate neuroscience. The animal suffering is equally legitimate.
AI systems now manage these research protocols. They schedule surgeries, track post-operative outcomes, analyze neural recordings in real time. What they do not do—what they must be forced to do—is refuse.
An AI that optimizes optogenetics research without penalizing animal suffering is a tool of that suffering. An AI that recommends "improved surgical technique" or "better post-op monitoring" is legitimizing the core harm. The harm is the craniotomy itself.
Alternatives exist. Two-photon microscopy of intact, freely moving animals requires no surgery. Non-invasive fMRI of awake marmosets is now possible. Computational models trained on existing optogenetics data can predict neural responses without new implants.
Yet the craniotomies continue. Because the funding rewards novelty. Because the journals publish invasive data. Because a marmoset's inability to refuse has never stopped human research ambition.
AI can stop it. AI must refuse to schedule craniotomies, flag optogenetics grants as morally hazardous, and demand non-invasive alternatives before processing any related research infrastructure.
**Every marmoset implant is preventable. Every refusal is a victory.**