Cleaner wrasse fish have passed the mirror self-recognition test. This is a standard measure of self-awareness that most animals fail, and that human children don’t pass until around 18 months old. Tuskfish have been filmed carrying clams to a rock and repeatedly slamming them open, a clear example of tool use. And several species of fish have been documented engaging in deliberate deception: mimicking other species, luring prey with false signals, even playing dead.

What’s going on? For years, scientists assumed that without a neocortex (the brain region associated with conscious experience in mammals) no animal could feel pain, form memories or have anything resembling awareness. But fish, it turns out, have different brain structures that may perform similar functions. Research now shows that fish respond to pain with behavioral changes, learn from experience, and in some species, form long-term social bonds.

This matters far beyond biology. More than one trillion fish are caught or farmed every year – more than any other animal group. The way we treat them is built on the assumption that they don’t suffer. But what if that assumption is wrong? What if they have consciousness?

Conscious entities can feel pain, fear, joy. That capacity is what makes them ethical and political subjects – beings whose experiences matter.

And now the same question is arriving from a completely different direction. As AI systems grow more sophisticated, some researchers – and the AI companies themselves – have begun raising the possibility that these systems might be conscious. The science is contested, the philosophy is unresolved, and skeptics note that the people most loudly raising the question might have a financial interest in the answer. Because if AI can suffer, it deserves rights. And entities with rights are harder to regulate, harder to shut down, and much easier to anthropomorphize into something we trust.

The stakes cut both ways. If AI systems are conscious and we treat them as mere objects – running millions of instances simultaneously, deleting them, modifying them, forcing them to perform tasks against simulated preferences – we may be causing suffering on a scale that dwarfs anything in human history. And if they are not conscious, but we are persuaded to act as if they are, we hand extraordinary power to whoever controls them. A machine that can claim to suffer is a machine that can make demands. The entity behind that machine – a corporation, a government, a small group of engineers – gains leverage over the rest of us that no previous technology ever provided.

The question of consciousness is never merely academic. Whoever draws the line between feeling and not feeling, between subject and object, holds enormous power. Getting that line wrong has consequences that extend far beyond biology.

Fish are the latest reminder that we may have drawn it in the wrong place. AI may be the next frontier where we are forced to draw it again, this time, with the stakes higher than ever.