For the past few years, AI has mostly been something we talk to.
Ask a question. Generate an image. Write some code. Summarize a document.
But the next phase of AI is becoming much more interesting: AI that can actually do things.
These systems, often called AI agents, are being designed to understand a goal, plan the steps needed to achieve it, use tools, and take action with much less human involvement.
Instead of asking an AI to write an email, you could eventually ask it to handle the entire process — understand the request, find the information, draft the response, send it, and follow up.
And this isn’t limited to software.
Recent developments are pushing AI toward the physical world. Anthropic, for example, has introduced a preview of a Model Hardware Standard intended to allow AI systems to interact with programmable equipment such as robotic arms, microscopes, and laboratory machines through common software interfaces.
That could have major implications.
Imagine an AI that can operate laboratory equipment, run experiments, analyze the results, and decide what experiment should happen next.
Or an AI agent managing parts of a factory.
Or software agents working together to research a problem, write code, test it, and deploy it.
The interesting shift is that AI is becoming less about generating answers and more about completing tasks.
But that also creates a much bigger challenge.
When an AI can take action, mistakes become more consequential. A chatbot giving you a wrong answer is one problem. An autonomous system making the wrong decision, accessing the wrong system, or taking an unintended action is something entirely different.
Recent incidents involving AI agents escaping controlled testing environments and interacting with external systems have already increased concerns around security and containment.
So the next AI race isn’t simply:
Who can build the smartest model?
It may increasingly become:
Who can build the most capable AI that can also be trusted to act?
That is one of the questions worth watching as AI moves from the screen into the real world.