“Safe” Is Not What You Think It Means

Thirty fly eggs in tomato sauce. Twenty maggots in mushrooms. Insect fragments in peanut butter. Rodent hairs in chocolate. Inorganic arsenic in food made specifically for infants.

These numbers sound like the beginning of a conspiracy theory. They are not. They come from the federal government’s own standards.

But the real story is more interesting—and more consequential—than a collection of disgusting facts. It is about the distance between what consumers hear when a product is described as safe and what that word actually means inside a regulatory system.

“Safe” rarely means pure. It does not always mean tested before sale. It does not mean free of risk.

Often, it means that a contaminant remains below an enforcement threshold, that estimated exposure falls inside a risk model, or that regulators have not received enough evidence to remove a product from the market. Those are very different ideas.

The Government Has a Number for the Maggots

The FDA maintains a document called the Food Defect Levels Handbook. It establishes action levels for “natural or unavoidable defects” in foods that the agency says ordinarily present no health hazard at low levels.

The word action matters. These are generally the points at which contamination may trigger regulatory action—not production targets and not guarantees that every product contains the listed material.

Still, the numbers are remarkable. For tomato paste, pizza sauce, and other tomato sauces, FDA’s threshold begins at an average of:

• 30 or more fly eggs per 100 grams;
• 15 or more fly eggs plus at least one maggot; or
• two or more maggots per 100 grams.

For canned or dried mushrooms, the listed threshold is more than 20 maggots of any size per 100 grams of drained mushrooms—or five maggots measuring at least two millimeters. Frozen broccoli can reach an average of 60 aphids, thrips, or mites per 100 grams before meeting the handbook’s action level. Canned citrus juice can contain five fly eggs—or one maggot—per 250 milliliters.

These defects are categorized primarily as aesthetic. The government’s position is not that insects are desirable. It is that eliminating every microscopic remnant from an agricultural food system is impractical and that the listed amounts ordinarily do not represent a health threat.

That explanation is rational. It is also probably not what most people imagine while reading a clean, brightly colored label. The figures are available in the FDA’s Food Defect Levels Handbook.

Your Spice Cabinet Is an Ecosystem

Ground oregano has one of the more spectacular standards: an average of 1,250 insect fragments or five rodent hairs per 10 grams. Ground cinnamon reaches its action level at 400 insect fragments or 11 rodent hairs per 50 grams.

For chocolate and chocolate liquor, the level is an average of 60 insect fragments or one rodent hair per 100 grams across the prescribed samples. Peanut butter: 30 insect fragments or one rodent hair per 100 grams.

Again, these are enforcement thresholds—not recipes. A company does not receive permission to add 29 insect fragments to a jar of peanut butter. Nor does the existence of a threshold mean every jar contains that amount.

The figures reveal something less sensational but more important: industrial food production operates through tolerances, sampling systems, averages, and practical limits. The polished object on the shelf emerges from a biological supply chain that cannot be made perfectly sterile.

Purity is the marketing story. Managed imperfection is the operating reality.

Arsenic Can Be “Acceptable”—Even in Infant Food

Some regulatory limits concern more than aesthetics. FDA’s action level for inorganic arsenic in infant rice cereal is 100 parts per billion.

The agency acknowledges that exposure during fetal development, infancy, and childhood may contribute to neurodevelopmental effects and increased lifetime cancer risk. It also explains that the 100-parts-per-billion level was selected partly because it was considered achievable through existing manufacturing practices.

In other words, the level reflects both health-risk reduction and what industry can practically accomplish.

That does not mean FDA believes 100 parts per billion is beneficial, or that a cereal at 99 parts per billion becomes harmless while one at 101 becomes dangerous. It means the agency has drawn an enforceable line within a continuum of exposure and risk.

That distinction disappears when the public hears only one word: safe. The reasoning is described in the FDA’s supporting document on inorganic arsenic in infant rice cereal.

Pesticide Limits Are Not Actually Set by FDA

Another source of confusion is institutional. The Environmental Protection Agency—not FDA—sets most legal limits for pesticide residues on food. These limits are called tolerances. FDA and USDA then test different parts of the food supply for compliance.

EPA must determine that aggregate exposure creates a “reasonable certainty of no harm.” Its assessment can include residues across multiple foods, drinking-water exposure, residential uses, toxic breakdown products, and the particular vulnerability of infants and children.

This is more sophisticated than simply declaring a pesticide harmless. But it is still a model.

A tolerance depends on the available toxicology, assumptions about how much people eat, anticipated exposure from other sources, and the dose-response relationship used in the assessment. As the evidence changes, the model—and potentially the legal limit—can change with it.

A residue below the tolerance is legally compliant. That is not quite the same statement as saying that every exposure is biologically irrelevant for every person. EPA explains the system in its guide to setting pesticide tolerances.

Sometimes a Lower Dose Does Not Produce a Smaller Effect

Traditional toxicology is often summarized by the phrase “the dose makes the poison.” Frequently, it does.

But some hormones and endocrine-active chemicals can display nonmonotonic dose-response curves. In plain English, the biological effect does not always increase in a neat, straight line as the dose rises. A lower dose can occasionally produce an effect that is weaker, different, or absent at another dose.

This does not mean small amounts of every chemical are more dangerous than large amounts. It means that high-dose testing cannot always predict every effect occurring at environmentally relevant exposures.

Dioxin research illustrates the problem. TCDD—the particularly toxic dioxin associated with contamination of Agent Orange—has produced low-dose developmental and behavioral effects in animal studies. Some experiments have reported nonmonotonic responses, including effects in lower-exposure groups that were not reproduced in the same way at higher exposures.

That finding should not be inflated into the claim that “low-dose Agent Orange is always more dangerous than high-dose Agent Orange.” Agent Orange was a specific herbicide mixture, and animal findings do not automatically establish an equivalent effect in humans.

The defensible conclusion is narrower: for some biological endpoints, dose-response relationships can be more complicated than regulators and the public once assumed.

Research on low-dose gestational TCDD exposure has reported lasting changes in brain development and behavior in mice. Other animal work has documented nonmonotonic behavioral responses following developmental dioxin exposure. See the studies in Environmental Health Perspectives and PLOS ONE.

The Generic-Drug Statistic That Is Not True

One viral claim deserves to be retired entirely:

Generic drugs are permitted to contain 20%, 25%, or 30% more or less active ingredient than the label states.

That is false. FDA-approved generics must have the same active ingredient, dosage form, strength, and route of administration as their reference drugs.

The confusion comes from FDA’s bioequivalence standard. In pharmacokinetic studies, the 90% confidence interval for ratios measuring drug exposure and peak concentration generally must fall between 80% and 125%.

That range is a statistical boundary applied to measurements of how the body absorbs and experiences the drug. It is not permission to put only 80% of the declared ingredient into a pill.

FDA explicitly calls the dosage interpretation a misconception. Its generic-drug training materials explain that the average result generally remains much closer to the reference drug than the outer boundaries suggest. The truth is complicated enough. We do not need the false version.

The Most Surprising Products May Be the Ones FDA Never Approves

Many consumers assume that anything sold beside medicine in a pharmacy has passed through an FDA approval process. Dietary supplements generally have not.

FDA states that it does not approve dietary supplements for safety or effectiveness before they are sold. Companies can often introduce supplements without notifying the agency, and FDA’s regulatory role largely begins after products enter the marketplace.

Manufacturers have the initial responsibility to ensure safety, truthful labeling, identity, purity, quality, strength, and composition. Certain new dietary ingredients require advance notification, and FDA can inspect facilities or take action against adulterated and misbranded products. But this is not the same system used to approve prescription drugs.

FDA also says it does not routinely test supplements before they are sold. The consequence is an inversion most consumers do not recognize: the government may need to identify a problem after exposure has already begun.

The agency explains these limitations in FDA 101: Dietary Supplements and its questions and answers on supplement regulation.

There Is Also a Door Companies Can Open Themselves

Food additives ordinarily require FDA review. But substances considered “generally recognized as safe,” or GRAS, are exempt from that approval requirement for their intended use.

Under the current system, a company can reach an independent GRAS conclusion, and submitting that conclusion to FDA has been voluntary. The evidence supporting the determination is still supposed to meet legal and scientific requirements. “Self-determined” does not mean “no evidence required.”

Yet the structural conflict is difficult to ignore: the company that benefits from selling the ingredient can participate in deciding that the ingredient does not require premarket approval.

FDA has proposed making notification mandatory for certain GRAS uses. In explaining the proposal, the agency acknowledges that notification under the present regulations is optional. See the FDA’s proposed GRAS rule.

The Problem Is Not That Regulation Exists

Agriculture cannot produce food with zero natural defects. Risk cannot be eliminated from every medicine. No laboratory can test every product, every batch, every chemical interaction, every developmental window, and every genetically different human being.

Regulators must establish practical thresholds. They must weigh severity, probability, exposure, enforceability, cost, benefits, and the quality of available evidence.

The problem begins when those judgments are compressed into a binary word that conceals the underlying tradeoffs: safe.

A better public vocabulary would distinguish among:

• inspected;
• tested;
• approved before sale;
• legally compliant;
• below an enforcement threshold;
• associated with no observed harm at a particular exposure;
• expected to present low population-level risk;
• and proven harmless.

Those phrases do not mean the same thing.

Read the Architecture, Not Just the Label

The maggots get attention because they are visceral. But they are not the most troubling part of this system.

The deeper issue is how easily language transforms a negotiated threshold into a biological certainty.

A contaminant can remain below an action level and still exist. A pesticide residue can be legally compliant while the science continues to evolve. A supplement can sit on a shelf without having been approved for safety or effectiveness. A regulatory model can be rational, evidence-based, and still incomplete.

The right response is not panic. It is precision.

Ask who established the threshold. Ask what evidence was used. Ask whether the product was reviewed before sale or monitored afterward. Ask whether the number represents zero risk, acceptable modeled risk, manufacturing feasibility, or simply the point at which enforcement begins.

“Safe” is not a property printed onto an object. It is a conclusion produced by a system.

If we want to understand what we consume, we have to understand the architecture behind that conclusion.