When one of the people at the forefront of artificial intelligence development argues that we should take a closer look at the speed of progress, it is worth paying attention.
Dario Amodei, co-founder and CEO of Anthropic, argues that the capabilities of modern AI systems are advancing so rapidly that safety mechanisms and our understanding of these systems risk falling behind.
What makes this particularly noteworthy is that this warning does not come from someone who fundamentally opposes artificial intelligence. It comes from a company that is actively driving its development.
And that raises a much bigger question for me:
What happens if AI keeps getting faster, while our ability to use it responsibly fails to keep pace?
Technology Scales Faster Than Responsibility
A new AI model is released. Within weeks, companies begin using it to write texts, analyze data, develop software or support decision-making processes.
Our structures do not evolve at the same speed.
Organizations need to clarify responsibilities. Employees need new skills. Leaders must decide which tasks and decisions can be delegated to AI systems – and which cannot. Legislation takes time. And as a society, we still need to determine what kinds of AI use we consider acceptable.
That is exactly why I like to talk about guardrails when discussing AI ethics.
Guardrails Do Not Stop Speed – They Make It Safer
A guardrail is not built to prevent speed altogether. It is placed where speed is possible, but where a mistake could have serious consequences.
Applied to artificial intelligence, this means asking certain questions much earlier:
Who is responsible when AI contributes to a decision? Where must humans retain the ability to intervene? How transparent does a system need to be? What level of risk are we willing to accept?
And perhaps most importantly:
Just because we can automate something, should we?
These questions already affect areas such as recruitment, performance evaluations, medical applications, credit decisions, education and the way we organize work.
We Should Not Only Ask What AI Can Do
Much of the current AI debate focuses on one question:
What will AI be able to do next?
We should be asking a second question with equal intensity:
Under what conditions do we want it to do these things?
For me, this is at the heart of AI ethics. It is not about preventing technological progress or treating every innovation primarily as a risk. It is about creating structures that allow us to use powerful technologies responsibly.
This includes clear accountability, opportunities for human intervention, transparency, independent reviews – and sometimes the conscious decision not to use a technological capability yet.
Responsibility Must Grow Alongside Technology
Technological progress does not automatically equal societal progress.
Perhaps, alongside increasingly powerful AI systems, we need a second form of innovation: innovation in how we organize responsibility.
In my view, the success of the AI era will not be determined solely by how intelligent our systems become.
It will also be determined by whether our sense of responsibility grows with them.
