A chef who crafts a delicacy for sale in the United States can choose from more than 10,000 food additives to garnish the dish. Of these chemicals, 43% are labelled ‘generally recognized as safe’ (GRAS) and need not be approved by the US Food and Drug Administration (FDA).

The system has weaknesses. A manufacturer is responsible for assessing whether an additive it has developed is GRAS. Once that is done, the manufacturer is asked — but not required — to notify the FDA. There are no data to evaluate compliance systematically, but the FDA found during a 2010 crackdown on caffeinated alcoholic drinks that four out of four manufacturers queried had not done the required checks.

Even when manufacturers do submit GRAS determinations, there are concerns about the quality of the assessment. An ongoing project at the Pew Charitable Trusts in Washington DC reveals discomfiting gaps in the data. A search of three toxicological databases, including that of the FDA, showed that fewer than 38% of GRAS claims were backed up by FDA-recommended toxicology studies in animals (T. G. Neltneret al.Reprod.Toxicol.http://dx.doi.org/10.1016/j.reprotox.2013.07.023;2013). The same team has published an analysis of 451 GRAS notifications submitted to the FDA. To avoid conflicts of interest, assessments should be done by an independent expert panel, but none had been; in fact, 22% had been performed by an employee of the manufacturer (T. G. Neltneret al.JAMAIntern.Med.http://doi.org/nd5;2013).

The FDA told Nature this week that it plans to issue guidance on how to fulfil GRAS requirements. One suggestion, proposed by Pew, is to take conflict-of-interest policies used to select FDA advisers, and apply them to GRAS determinations. The FDA should seize the opportunity to protect public health and boost confidence.