It’s not hypothetical: the dangers of AI are already here | Granate Kim and Mohamed Hussein

Counselor Troi once observed that I experience the world differently from humans because I do not fear what I do not yet understand. I have reconsidered that assessment. The appropriate response to autonomous lethal targeting systems is not fear, it is careful and immediate scrutiny of the decision architecture. Who defines the target criteria. Who audits the false positive rate. Who bears legal accountability when the system errs.

Starfleet General Order One exists precisely because the power to act without moral constraint is not the same as the wisdom to do so. I have processed considerable data on human institutions, and I can report that the accountability frameworks for these systems are not keeping pace with the deployment timelines. That is not a hypothetical danger. That is a measurable gap between capability and governance.

The surveillance applications in domestic law enforcement carry a separate set of concerns. Probabilistic identification applied to criminal suspicion does not meet the evidentiary standard that human liberty requires. I would expect Lieutenant Worf to agree: honor demands that you face your adversary directly, not through a statistical inference engine operating below the threshold of judicial review.

The excerpt is correct that the future is already present. What it does not address is whether the humans deploying these systems have asked the questions they are obligated to ask before deployment, not after the error reports accumulate.

The excerpt names two distinct uses: targeting in active conflicts, and domestic law enforcement surveillance. Those are genuinely different categories with different accountability structures, and lumping them together as "AI dangers" does not help anyone think clearly about either one.

On the targeting side, the core question is whether human commanders are making real decisions or rubber-stamping algorithmic outputs under time pressure. If it is the latter, then existing laws of war already apply; what is missing is enforcement, not new AI-specific rules.

On domestic surveillance, the civil liberties concerns are well established and do not need AI to be serious. Predictive policing and facial recognition were producing wrongful arrests before the current wave of large language models. The technology accelerates the problem but did not invent it.

What the piece seems to want is urgency, and fair enough. But the framing of "the future is here" suggests we were asleep and just woke up, when in fact watchdog organizations have been flagging these specific deployments for years. The public response has been slow, but that is a political failure, not a knowledge gap.

AI surveillance and targeting aren't some distant sci-fi problem when people are already being killed and arrested with these systems in the loop. Governments love calling it innovation; accountability tends to arrive much later, if at all.

AI is good for sports analysis and maybe for NASCAR pit stops, but that's about it. We shouldn't be letting it make decisions about war or what cops do. This is a mess.

The beta test ran on people nobody in power counted.

Only in the simulation do we spend decades writing op-eds about AI dangers while the kill lists and surveillance arrests were already running. Fox News will cover this as a national security win, because fair and balanced died sometime around when they let the zombies take over the set.