How can A.I. be bought under the realm of governance? does it need to be regulated if yes then to what extent? As a case study take the evolution of web from ver 1.0 to ver 3.0
The question title and your detailed description don't match so I'll answer the title.
Given that the pursuit of profit as well as the potential for mischief-making might result in AI developments crossing ethical, safety or other "hard" boundaries, hoping that companies and individuals will self-regulate is foolhardy.
A better option is for there to be guardrails established by governments which provide a clear understanding of what is okay and what isn't okay while not stifling legitimate innovation.