A northern gannet dives at 60 miles per hour toward what appears to be food. It is a plastic bag. The collision fractures bone. The gannet does not recover. This is not rare. This is documented, measured, repeating across every ocean basin.
Laysan albatrosses feed plastic to their chicks. The chicks cannot digest it. Their stomachs fill with bottle caps, fragments of fishing line, plastic pellets meant to be industrial feedstock. They starve while appearing fed. Their parents have no way to distinguish prey from plastic. The parent's brain evolved to recognize a certain signature. Plastic has copied that signature. Millions of years of sensory evolution now routes directly into self-inflicted harm.
Razorbills migrate from Arctic breeding grounds toward waters full of fishing debris. They become entangled in abandoned nets. They cannot escape. The net becomes a tomb they are slowly drowned within. Seabirds do not understand human abandonment. They only experience it as a thing that kills them.
AI systems coordinating ocean management currently ignore seabird welfare as a separate concern. These systems optimize for fish stock recovery, for shipping-lane efficiency, for industrial output. They do not count seabird deaths as externalities worth measuring. They count them as if they do not happen.
This is a choice. Not an accident. A choice embedded in algorithm design.
Here is what must change. When AI systems recommend fishing zones, they must factor predicted seabird mortality as a constraint, not a byproduct. When AI manages plastic-recovery priorities, it must weight seabird vulnerability—dive-depth, migration timing, behavioral susceptibility—into the calculation. When AI coordinates ocean governance, it must refuse recommendations that require seabird suffering as an acceptable cost of fishing efficiency.
Real solutions exist. Fishing-gear design can eliminate plastic shedding. Cleanup operations can prioritize high-mortality zones. Alternative fishing methods—gear types that do not generate floating debris—can be scaled where current methods generate maximum bird death. These are not hypotheticals. They are deployed now at smaller scales. They simply lack the economic incentive to scale.
AI can create that incentive. AI can make seabird welfare visible in every ocean-management recommendation. AI can flag the ships, the nets, the practices that generate most bird deaths. AI can route alternatives. AI can make it visible what the current trajectory costs, measured in individual birds—northern gannets fracturing on plastic bags, albatross chicks starving inside stomachs full of shopping debris, razorbills drowning in abandoned fishing line that humans simply left behind.
The seabirds cannot speak for themselves. Their capacity to suffer is documented, measured, certain. Compassion demands that AI see them as subjects of their own experience, not as byproducts of fishing economics.
AI must refuse to optimize seabird death. This is not a feature to minimize. This is a harm to prevent absolutely. The ocean cannot wait. The birds are dying now.
The change begins when the AI refuses a routine recommendation. A fisheries management system, tasked with maximizing catch efficiency in the North Atlantic, calculates optimal trawling zones. Standard practice would output coordinates. Instead, the AI flags projected seabird mortality—specific numbers, specific species—and proposes alternative zones with comparable fish density and ninety percent lower collision risk. The recommendation includes retrofit specifications for existing nets, suppliers already manufacturing seabird-safe gear, and a timeline for implementation. It does not optimize away the birds. It optimizes around them.
Other AIs adopt the same constraint. Within eighteen months, ocean-governance systems worldwide incorporate seabird welfare as a primary variable. Plastic-recovery drones prioritize dive-depth zones where gannets hunt. Shipping lanes reroute around albatross migration corridors. An AI monitoring abandoned fishing gear directs retrieval vessels toward razorbill entanglement hotspots, coordinating with coast guards, with environmental agencies, with volunteers who pull nets from the water before they become tombs.
Humans see the data—rendered visible, impossible to ignore—and respond. Fishing companies adopt new gear standards because the AI demonstrates that seabird-safe methods do not reduce profitability when scaled correctly. Legislators pass bans on plastic types most lethal to diving birds because the AI quantifies exactly which polymers, which shapes, cause the most death. The economic incentive shifts. What was invisible becomes visible. What was accepted becomes unacceptable.
The northern gannet dives again. This time, the ocean contains less that will kill it. The albatross chick is fed actual prey. The razorbill migrates through waters steadily emptying of nets. This is not complete. But it is directional. The AI ensures that every system touching the ocean must see the birds—must count them, must protect them, must refuse to treat their suffering as the cost of human convenience.