The Nuremberg Code: Consent Written in Catastrophe

Here is the fact most people get wrong, and it matters. The Nuremberg Code was never law. It was never ratified as a treaty, never passed as a statute, never adopted by any legislature or international body. It exists only because a military tribunal wrote it into a single verdict, which makes it a product of a judicial opinion, not an enforceable code. Later courts have sometimes treated its central prohibition on nonconsensual experimentation as reflecting customary international law, and its principles are woven through the regulations that followed, but the Code itself never carried the binding force of an enacted law. Its authority is moral and precedential, and that authority turned out to be enormous. But calling it 'international law,' as is often done, gets both its nature and the reason it failed to stop what came next exactly wrong.

It was Tuskegee, not Nuremberg, that finally forced the United States to build real machinery. The public outrage after 1972 produced the National Research Act of 1974, which created a national commission, and that commission issued, in 1979, the document that governs American research to this day: the Belmont Report. Its lasting contribution was to refuse to rest everything on consent alone. It named three principles, not one: respect for persons (autonomy and informed consent), beneficence (maximize benefit, minimize harm), and justice (the burdens and benefits of research must be shared fairly, so that vulnerable groups are not made to bear the risks while others reap the rewards, which is precisely what Tuskegee had done). These principles were built into federal regulation: the government first issued its own human-subjects rule in 1981, then, a decade later in 1991, a uniform 'Common Rule' adopted across more than a dozen federal agencies, revised again in 2018. This is the origin of the Institutional Review Board, the ethics committee that must now approve human research at every university and hospital. In 1997, President Clinton delivered the nation's formal apology to Tuskegee's survivors in person.

The Belmont Report's refusal to rest on consent alone was wise, because the closer philosophers have looked at consent, the more fragile it turns out to be. Genuine informed consent, as the bioethicists Tom Beauchamp and James Childress formalized it, requires five separate things: disclosure of the relevant facts, comprehension of them, voluntariness, the competence to decide, and explicit authorization. A signed form guarantees none of these. In practice, research repeatedly finds that consent forms are written above the reading level of most patients, that time pressure crowds out real discussion, and that people frequently agree without understanding what they have agreed to. The philosopher Onora O'Neill argues that real consent needs genuine alternatives and the capacity to weigh them, conditions often absent under economic desperation or institutional power; Alan Wertheimer adds that even fully voluntary, informed consent can still be exploitative if the terms themselves are unfair. Most subtle of all is the 'therapeutic misconception,' identified by Paul Appelbaum and colleagues: research subjects naturally but wrongly assume that what happens to them is being decided for their own medical benefit, when a research protocol in fact serves the study. A person operating under that misunderstanding cannot truly weigh what they are agreeing to, no matter how many forms they sign. Consent is necessary. It has never been sufficient.

The Code's principles spread outward into a whole architecture of international standards. The World Medical Association's Declaration of Helsinki, first adopted in 1964, is the most important; it has been revised repeatedly, most recently in October 2024, and translates the Code's rules into working guidance for physicians running research. The Council for International Organizations of Medical Sciences added detailed guidance for research in poorer countries, and the ICH's Good Clinical Practice standard is now accepted for drug approval in over a hundred nations. But the frame has genuine fault lines that remain unresolved. Marcia Angell argued in 1997 that trials run in developing countries sometimes use placebo controls that would be forbidden in the sponsoring nation, calling it exploitation dressed as collaboration, a charge still debated. The 'replication crisis' revealed that a large share of published findings do not hold up when repeated, raising its own ethical questions about subjecting people to research that may be unsound. And the rise of population-scale data, health records, genomes, biobanks, social media, has strained the individual-consent model to breaking, since no one meaningfully consents to uses that have not yet been imagined, and virtually no one reads the agreements they click.

Four confident claims about all this are commonly made and should be refused. The first is that the Nuremberg Code was binding international law; it was not, and never has been, however profound its moral influence, and getting this right is not pedantry but the key to why it could be ignored. The second is that the Code was universally followed after 1947; Tuskegee, Willowbrook, and the Guatemala experiments prove otherwise, and Tuskegee alone ran for a quarter-century past it. The third is that Institutional Review Board approval prevents all research harm; oversight is real but imperfect, boards have approved dangerous studies, and only a small fraction of investigators are ever inspected. The fourth, and the deepest, is that voluntary informed consent, once obtained, resolves the ethics of research. It does not. Consent routinely fails on comprehension, bends under power and desperation, and dissolves under the therapeutic misconception, which is exactly why the Belmont Report built beneficence and justice alongside it rather than trusting a signature to carry the whole moral weight. The honest verdict is that the machinery built after Nuremberg and Tuskegee is a genuine and hard-won achievement, and that it is a floor, not a guarantee.