Anyone who has helped to run an organization knows that good governance includes regular evaluation of risks — risks to employees, to security, to processes, to reputation, to financial viability — and, where these risks are considered excessive, actions to mitigate them. Risk assessments of this type need to be specific, so inevitably they are confidential.

Research funders who steward the allocation of millions or billions of dollars routinely assess and manage such risks. But they also deal with other kinds of risk — and success in those areas should not be kept quiet.

There are two types of risk management that funders should highlight: one that deals with bad risk, and the other with good.

The bad risks involve funds being wasted on inept, sloppy or downright fraudulent research. Funding agencies are usually responsible to taxpayers; it is therefore crucial that they are publicly accountable for ensuring that they support robust science. This is easily said. How can it be done? The least that the funder can do is to ensure that the institutions that it funds have procedures for tackling sloppy or dishonest research when it arises and, better still, for reducing the likelihood of such problems through appropriate scrutiny and training.

Journals tackling the challenges of reproducibility have found that measures to ensure that research is robust can be onerous for editors, referees and authors. The same will be true for funders. But if public trust in science is to be maintained, funders need to mitigate bad risks through assessment procedures. They could, for example, ensure that claims about the statistical power of proposed experiments stand up. What is more, they need to be seen to be doing this.

One funder that has taken these responsibilities seriously is the US National Institutes of Health. To tackle poor reproducibility in biomedical research, it has arranged discussion meetings, introduced training programmes and made its assessment of grant applications more rigorous. It has also made these initiatives highly visible: see go.nature.com/8jhvwq. Others should follow its example.

Then there are the good risks. At a meeting of the Global Research Council in Tokyo last month, funders produced a declaration of principles to address the challenges of funding scientific breakthroughs (see go.nature.com/nrlxmd). The most consistent point was that funders should feel able to take positive risks — that is, to support daring ideas that have a chance of failure. They should “encourage risk-taking and tolerate failure in research activities”. And they should “provide researchers with the flexibility and intellectual space needed for serendipity”. Much of this comes down to funders ensuring “the freedom of researchers in defining their topics, their methodologies and their resource allocations”.

Although such boldness is not easy to implement, researchers know what the combination of great talent and bold ideas can deliver. Outsiders, including governments and taxpayers, might not appreciate the potential value of positive risk-taking. But it goes with the territory of supporting outstanding science.

It is important, therefore, that funding-agency managers empower selection committees to accept a degree of risk in supporting boldness in research, and to back researchers up when (and it will be when, not if) some of those risks do not pay off. Any funding agency that lives up to these principles should proclaim its risk-taking mission with pride.