The Speed Problem

The rapid advance of artificial intelligence has outpaced our collective ability to think clearly about its consequences. This isn’t because the risks are obscure, but because we’ve grown accustomed to treating powerful technologies as neutral tools rather than as systems that reshape incentives, behavior, and moral responsibility. The recent controversy surrounding Grok makes this failure hard to ignore.

What’s revealing isn’t that a generative model produced harmful output. Any sufficiently complex system will fail in predictable ways. The problem is that these failures now occur at scale, in public, and faster than any serious mechanism of oversight can respond. Emergency bans, geoblocking, and reactive policy updates aren’t evidence of control – they’re evidence of panic.

We continue to deploy AI systems as if they were consumer novelties, while they function more like infrastructure. They influence what people see, believe, and create, yet they are governed with the same casualness as social media features. When harm emerges, responsibility diffuses instantly – across developers, platforms, regulators, and users – until it effectively disappears.

The speed of progress is the central danger. Each iteration is more capable, more autonomous, and more embedded in everyday life. Meanwhile, our legal frameworks, ethical intuitions, and institutions update at a human pace. This asymmetry matters. A system that can generate convincing images, narratives, or misinformation in seconds can do real damage long before society agrees on how such damage should even be defined.

There’s also a persistent confusion about where the problem lies. We fixate on individual products – this model, that scandal – as though removing one instance addresses the underlying issue. It doesn’t. The trajectory is what matters. Capability is compounding, while judgment is not.

None of this requires alarmism. It does require intellectual honesty. AI is not slow, and it is not neutral. Treating it as either guarantees that its failures will continue to surprise us, even as they become increasingly predictable. The Grok episode is not an anomaly. It is an early signal of what happens when powerful systems are released faster than we are willing to think about what should constrain them.

The question we keep avoiding is the only one that matters – not what these systems can do, but what kinds of systems should exist at all, and under what limits. Until that question is taken seriously, we will remain stuck reacting to consequences that were obvious in advance.