We have known for a long time that many safety features do not actually make people safer because they use this opportunity to work faster/riskier/etc and thus revert to the individual's own risk tolerance (modified by external incentives such as productivity bonuses).
But there are also negative externalities to worry about. In a new study, Ian Walker at the University of Bath found that cyclists who wear helmets have a higher chance of getting hit by a car than those who do not. So does this mean that cyclists adjust to their added safety by riding more aggressively? It turns out the answer is no. The people who adjust their behavior is the driver. Drivers passing cyclists with helmets drive closer to the bike than drivers passing cyclists without helmets. So even though the biker doesn't change his/her behavior, his/her risk goes up.
An earlier study by Sam Pelzman found that drivers who wear seat belts drive more aggressively. So their risk stays the same overall. But the people around them have higher risk. So the total system risk is higher. Paradoxically, adding safety features increases total system risk.
There is not enough data to know how significant this effect is in general or whether it would really be better to get rid of some safety features to increase safety. Sounds like a good research proposal for an ambitious safety researcher.