The disagreement about assurances is not between people who oppose torture and people who do not. It is between people who think a promise from a torturing state can be made reliable enough to act on, and people who think the attempt corrupts the prohibition it is meant to satisfy. Both positions are held by serious scholars, and both are set out here in their strongest form, with a link to the writer making the argument.
The case for allowing assurances
The strongest statement of this position is Ashley Deeks, then writing for the Council on Foreign Relations. Deeks acknowledges the critics’ central point, that assurances from unreliable states cannot be enforced and are difficult to monitor. Her argument is that the practice survives because the alternatives are worse. Prosecution in the sending state is frequently impossible for evidentiary reasons. Indefinite detention is undesirable and, in many systems, unlawful. Release is unacceptable to the sending government. Against that set of options, she argues, an assurance backed by stronger oversight and by practical assistance to the receiving state’s prison system is the least bad course.
Ryan Goodman reaches a related conclusion by a different route. He rejects the categorical position on both sides: the United Nations special rapporteurs are wrong to reject assurances outright, and the United States is wrong to apply a “more likely than not” threshold. He argues that the Human Rights Committee’s sliding scale, which weighs how systematic torture is in the receiving state and requires rigorous monitoring, better reflects the international consensus. Note the sting in his conclusion: because that approach assumes a “real risk” threshold stricter than the American standard, the United States cannot borrow it to justify its own practice.
The Strasbourg case law points the same way as Goodman rather than as Deeks. Aristi Volou, writing on Article 3 of the European Convention, describes a pragmatic, quality-of-assurances analysis under which assurances can be adequate guarantees of safety in certain circumstances. The European position, in other words, is neither prohibition nor deference: it is scrutiny of the particular assurance.
The case against
Gregor Noll makes the structural argument. Assurances are in substance bilateral agreements about a person, and what they agree on is a disagreement: the receiving state does not concede that it tortures, and the sending state does not press the point. Noll argues that they silence human rights law by moving supervision out of courts and into diplomatic networks, that they leave the person unable to report abuse safely because the only channel runs through the government holding him, and that their function is to protect the sending state’s self-image rather than the person’s body.
The treaty bodies have taken a hard line. In their survey of more than five hundred non-refoulement decisions of the United Nations treaty bodies, Başak Çalı, Cathryn Costello and Stewart Cunningham report that the Committee Against Torture has been highly resistant to assurances, treating them as loopholes. The authoritative commentary on the Convention, by Margit Ammer and Andrea Schuechner, concludes that assurances cannot substitute for a genuine assessment of risk.
The American procedural critique is narrower and, for this site’s purposes, more immediately useful. Brenna Nelinson argues that the federal courts have improperly foreclosed review of torture-risk determinations by deferring wholly to executive assurances, and that meaningful review is what the Convention and the implementing statute require. Steve Vladeck, writing on a Fourth Circuit appeal involving three Rwandan nationals, frames the question as whether assurances can ever justify removal into a torture risk, and identifies Khouzam as the source of the rule that a person must at least have the chance to test an assurance’s reliability.
Where the two sides agree
More than the argument’s temperature suggests. Almost nobody defends an unmonitored assurance accepted on the executive’s word alone and unreviewable anywhere. The Columbia Law School Human Rights Institute study, which declines to take a position for or against assurances as such, identifies the conditions any credible practice would have to meet: judicial review, public scrutiny, and systematic monitoring. Deeks argues for stronger oversight. Goodman argues for rigorous monitoring. Nelinson and Vladeck argue for judicial review. The disagreement is about whether those conditions can in fact be met, not about whether they are the right conditions.
That is worth holding onto, because it identifies what to look at in any particular case. An assurance that is written down, disclosed to the person, open to challenge before a neutral decision-maker, and backed by monitoring that the person can trigger is a different instrument from one that is none of those things, even though both are called by the same name.
The current program
Assurances have become material again because they are being used again. Scott Roehm and Rebecca Ingber, writing on six congressional resolutions concerning removals to Costa Rica, Mexico, Panama, Rwanda, South Sudan and Eswatini, argue that reliance on diplomatic assurances cannot safeguard against torture, and point to El Salvador, where a written assurance was given and abuses followed. Their second point is the one that connects this project to the Coalition’s other work: the resolutions ask whether receiving states may remove transferees onward. An assurance addresses what the receiving state will do. It does not address what a fourth state will do after the receiving state sends the person there.
Two recent treatments locate the same problem inside the removal process. Megan Elman-Welch, writing in the Georgetown Immigration Law Journal, describes how countries of removal are designated in immigration court in seconds at a preliminary hearing and calls assurances unreliable. Matthew Boaz argues that people who hold protection should receive meaningful administrative review before removal to an unconnected country, and that the assessment must include a serious inquiry into whether onward return is a risk.
The sources for every position described on this page are listed, with links, in the sources section.