“Game software might evolve not just to be more addictive, but to be safer, since killing the customer is counteproductive.”—mtraven
Argument from group selection. Killing the customer is bad for the industry, but not for the company. If everyone plays the most fun game on the market, and 10% of its players die annually from playing it, and you come up with a new, more entertaining game that will give you the entire market but with a 15% annual death rate, you don’t get rich by trying to save lives.
Also, an occasional “died playing” story might signal “really entrancing game” and work as advertising.
On the other hand, at least some of the market are likely to be deterrred by such stories. (The risk of government intervention also increases, but that’s bad for the industry not the company)
If companies had legal liability for deaths stochastically caused by their games, there would be negative feedback pressure on the individual company.