This is an open question, and it is open at both ends: we do not have a settled scientific account of human consciousness, so we have no reliable way to detect it in a machine. A system can be highly capable without being conscious, and the reverse is at least conceivable.
Why behaviour cannot settle it
Any behavioural test can in principle be passed by a system that simulates the behaviour without the experience. That is the core of the problem. Reports of inner states are also unreliable evidence, because a language model is trained to produce them.
The leading theories
Global workspace theory, integrated information theory and higher-order theories each imply different things about machines. Some suggest consciousness requires specific architectural features that current systems lack. None commands consensus, and several are contested even for biological brains.
Why it matters practically
If there is any realistic chance that a system has morally relevant experience, the ethics of training and deploying it change. That is a reason for caution, not a reason to assume consciousness has been achieved.