The term “generally recognized as safe”? You can now generally recognize as sound

It sounds simple, but it wasn’t. There was no way that FDA could rule on every single substance that manufacturers were already using in food, much less every new substance they were likely to come up with. So the agency grandfathered in few hundred ingredients and started making distinctions. Pesticides weren’t food additives, because they were already covered by a different piece of legislation. Ditto for food colorings. On the other hand, substances used in manufacturing equipment or packaging that transferred to food were additives.

And then came the big one: An ingredient in food was not to be considered an additive—and didn’t require pre-marketing approval—if it was generally recognized as safe, which the amendment defined as “generally recognized, among experts qualified by scientific training and experience to evaluate [their] safety, as having been adequately shown through scientific procedures (or, in the case of a substance used in food prior to January 1, 1958, through either scientific procedures or experience based on common use in food) to be safe under the conditions of its intended use.” (21 U.S.C. §321)

The FDA’s web page on GRAS quotes then-commissioner George P. Larrick explaining the intent of the language: “We believe only those chemicals should be automatically exempted from the new law which are recognized among competent experts as safe for their intended use. This would make it unnecessary, for example, to do studies on table salt, but would not approve the continued use, without proof of safety, of the synthetic emulsifiers now widely used in some fabricated foods.”

So how do you establish that something is GRAS? Life is too short for me to review the entire history of how FDA applied the concept. (The younger among you, or those confident of their longevity can indulge themselves in this fine article by attorney George G. Misko.) The short version of the story is that FDA was hesitant to nail down a procedure for a long, long time. It issued several preliminary rules, then failed to finalize them. (That’s a more common occurrence than you might think.) For a while, it had manufacturers file notices declaring their new ingredients were GRAS, and FDA would rule on them. Then it asked (but did not require them) to file, but mostly didn’t respond unless asked. Then it asked for filings, but never said that it approved or disapproved: It answered either that it didn’t have any questions, or that it didn’t have enough information. (The latter is the kind of notification Impossible Foods got for its soy leghemoglobin.) In any case, you could go on selling your product until FDA actively intervened and said you couldn’t.

The new guidance, if I may paraphrase rather freely, says, “Yes, that’s really what we meant. Carry on.” Manufacturers are still encouraged, but not required, to notify FDA that they have determined that their product is generally recognized as safe. They are reminded that the evidence they present has to be as scientifically strong as they’d need if they were filing for pre-marketing approval. And they’re warned that if they haven’t done the research, or if FDA decides that new evidence has emerged that the product is not safe, the agency can simply pull the plug on an ingredient.

As so often happens, the interesting stuff in the document is in the appendix, where FDA answers comments it received during its review process. Most of them focus on the issue of whether proprietary information can be used to determine that an ingredient is GRAS. And FDA is holding the line quite firmly. If you want to go the GRAS route for your product instead of the slower, more expensive pre-marketing approval route, you have to use publicly available, preferably peer-reviewed data. On the other hand, FDA also fends off the suggestion that “generally recognized” means there has to be unanimity in the expert community.

I know. The whole process sounds goofy. But when you stop and think about it, it’s not necessarily the worst way to regulate food additives. (I’m going to temporarily, respectfully, leave aside the point of view that all additives are intrinsically bad. It may be true, but it’s not a practical alternative at the moment.) Consider:

FDA has always been short on resources. I think we really do have to choose between pre-marketing approval and the sort of inspections and enforcement that have improved the safety of our food supply so much. The requirement that manufacturers self-certify will keep most potentially dangerous substances out of the marketplace—and really, even FDA’s high-powered, elaborate drug approval system doesn’t do much better than that. The dangers that sneak through the process will likely be subtle, emerging only after the product is on the market for a while.

And FDA isn’t the only party policing food manufacturers. There are plenty of activists and profit-oriented litigators out there ready to pounce on the slightest hint of impropriety. From that perspective, a system that lacks a formal approval stage robs manufacturers of a crucial defense they can use in lawsuits: the ability to argue “FDA said it was safe.”

And the emphasis on using published information is important as well. It means that those activists and litigators can look at the same material that was used in making the GRAS claim. There can be none of the secrecy that mars drug or ingredient approvals based on proprietary data. (Yes, companies can submit private data, but it can only be corroboration. The data used in deciding that something is GRAS has to be exclusively public.) That’s a big step toward transparency.

It’s true that companies don’t have to go through the GRAS process. They can petition for pre-marketing approval and back up their safety claims with proprietary data. But last time anyone looked, pre-marketing approval for a food ingredient at FDA took 72 months. Almost no one ever chooses it. Almost no one ever will.

It’s not a pretty system. It lets market forces do some of the things you might expect to be handled by government authority, which is something that hasn’t always worked well. (Remember the financial crisis?) It implicitly sets up some powerful disincentives for bad behavior, but kind of leaves it up to third parties to uncover malefactors. (Which can work, though often very slowly. Think Harvey Weinstein.) But the fact is that with a bit of nudging from interested parties, it’s likely to work as well as a more regimented system—at least until it doesn’t.

So, on balance, you can probably be reasonably happy with the new guidance. It’s not perfect, but in today’s climate of both technical innovation and extreme distrust, we aren’t likely to get better. Sleep easy.

But keep reading those labels. Your mama didn’t raise no idiots.