Perverse incentives can work in any direction you choose. I was talking here the other week about lab safety, and how it's a good thing to know where the fire extinguishers are. But what if you're working in a place where discharging one of those extinguishers sets off an avalanche of paperwork and committee meetings? Do you use the thing, or does the vision of all that wasted time give you pause, while the flames leap around your glassware?
If it's a seriously nasty fire, you're probably going to pull the pin and worry about the consequences later (and for a fire like that, it's good to remember that going for that second extinguisher is usually a bad idea, compared to, say, diving for the stairs). But what if it's just medium bad, and if you're not sure if it's going to get worse? Other things being equal, you should probably do the most effective thing you can to put it out. But other things aren't always equal in industry.
I've worked where the safety culture was limited to occasional warnings not to blow yourself up, and I've worked under intrusive, no-sparrow-shall-fall regimes. Neither of those, as far as I could see, kept me safer than the other. The problem is, if you're going to aggressively document every possible incident and near miss, to be entered into the massive database and discussed in detail at the mandatory regular safety meetings (attendance taken and computed into the year-end bonus formula). . .well, people are going to sit on most of the ones that they think that they can get away with. The harder you work to log every lapse, the more of them you'll miss.
Once people have reached a certain level of competence and experience, lab safety is largely a matter of thinking about what you're doing, realizing what you know and what you don't know, and planning ahead. These are all highly desirable qualities, both in and out of the lab, and they cannot be expressed by decree. No safety committee is going to make people smarter, and no multi-page web form will make them more alert. In this world, actually, the opposite is much more likely. . .