Incentives to innovate and social harm: laissez-faire, authorization or penalties?

When firms' research can lead to potentially harmful innovations, public intervention may thwart their incentives to undertake research by reducing its expected profitability (average deterrence) and may guide the use of innovation (marginal deterrence). We compare four policy regimes: laissez faire, ex-post penalties and two forms of authorization ??? lenient and strict. If fines are unbounded, laissez faire is optimal if the social harm from innovation is sufficiently unlikely; otherwise, regulation should impose increasing penalties as innovation becomes more dangerous. If fines are bounded by limited liability, for intermediate levels of expected social harm it is optimal to adopt (indifferently) penalties or lenient authorization, while strict authorization becomes optimal if social harm is sufficiently likely.