A promise by one government to another that a particular person will not be tortured after he is handed over. What the law requires before one may be relied on, what the courts have held, and why serious people disagree about whether the practice can be made safe.
A diplomatic assurance is a promise by one government to another that a particular person will not be tortured after he is handed over. It is the answer offered to an obstacle. The obstacle is that international law, and United States law implementing it, forbid sending a person to a country where there are substantial grounds for believing he would be in danger of torture. The assurance is meant to remove the grounds: the receiving state gives its word, the sending state accepts it, and the transfer proceeds.
The question the practice raises is not complicated to state. A country that would torture a person is a country whose promise not to torture him is worth asking about. Whether that promise can be relied on, who decides that it can, and whether the person it concerns ever gets to argue that it cannot, are the three questions this project follows.
Why this is a separate project
Assurances appear in three settings that are usually studied apart from one another: extradition, transfer from military custody, and immigration removal. The doctrine, the regulations, and the available review differ in each. A person removed from the United States on an assurance stands in a different position from a person extradited on one, and both stand in a different position from a person transferred abroad by the military. The law is more coherent when the three are read together, and the differences are easier to see than to explain when they are kept in separate files.
The practice is also older and wider than any single removal program. It has a body of case law reaching back two decades, a settled regulatory mechanism, and a substantial literature that divides sharply on whether assurances can ever be adequate. That combination is what makes a subject rather than a chapter.
What this project holds
Five sections. The first explains what an assurance is and how one is used, in plain words. The second sets out the treaty, the statute, and the two regulations that govern when an assurance may be relied on, quoting each as the courts have quoted it. The third reads the decided cases and reports what each court actually held, including where the courts have divided. The fourth sets out the argument as it is actually conducted, with the strongest case on each side and a link to the writer making it. The fifth is the source list, which is what makes the other four checkable.
Nothing here is legal advice. Nothing here tells any person what to do in any case. This is a record of what the law says and what the courts have done with it, published so that people can read the sources for themselves.
An assurance is a promise, given government to government, about how one person will be treated. It is usually short. It typically states that the named individual will not be subjected to torture, and it sometimes adds undertakings about where he will be held, whether he may be visited, and by whom. It is given in a diplomatic note or an exchange of letters rather than in a treaty, and it is not, in the ordinary case, published.
Where an assurance is used
Three situations account for nearly all of them.
Extradition. A foreign state asks the United States to surrender a person to stand trial. The person answers that he will be tortured if surrendered. The State Department seeks an assurance from the requesting state and, if it credits the assurance, the Secretary of State signs the surrender warrant.
Transfer from military or intelligence custody. A person held by the United States abroad is to be handed to the authorities of the country where he is held, or to a third state. The transfer is arranged with an assurance about his treatment afterward.
Immigration removal. This is the setting that matters most for the Coalition’s work. A person has already been found by an immigration judge to face a probability of torture, and has been granted protection for that reason. An assurance is then obtained from the very country the finding was about, and it is used to end the protection and permit the removal.
The structure of the problem
Read in sequence, the immigration case has a shape that is worth stating plainly, because it explains why the practice is contested rather than routine.
First, a tribunal makes a finding of fact: this person, if returned to this country, is more likely than not to be tortured. That finding is made on evidence, after a hearing, with the burden on the person to prove it. It is not lightly reached.
Second, the executive obtains a statement from the government that would do the torturing, to the effect that it will not. The statement is not evidence in the ordinary sense. It is not tested by cross-examination, it is not corroborated, and its author has an interest in the outcome.
Third, the executive weighs the statement against the finding and, if it credits the statement, the protection ends. Under the regulations, the decision that the assurance is sufficiently reliable is made by the Attorney General in consultation with the Secretary of State, and once it is made, the person’s claim for protection is not considered further by an immigration judge, the Board of Immigration Appeals, or an asylum officer.
The result is that a finding made after a hearing is displaced by a statement made without one. Whether the person may be heard on the reliability of that statement is the question the Third Circuit answered in 2008, and the answer it gave is set out in the section on what the courts have held.
What an assurance does not do
An assurance does not change the facts in the receiving country. It does not create a right that the person can enforce anywhere: he is not a party to it, and a promise between two governments about a third person is not ordinarily a contract that the third person can sue on. It does not, by itself, provide for monitoring, and where monitoring is promised, the promise runs to the sending state rather than to the person. And it does not address what happens if it is broken. There is no established remedy in the sending state for a person who has already been removed, and the practical difficulty of proving a breach from inside a prison in the receiving country is obvious.
None of that settles whether assurances are ever appropriate. Serious people argue that they are, on the ground that the alternatives are worse. That argument, and the answer to it, are set out in the section on what is disputed.
Four texts govern the use of an assurance in the United States: a treaty article, a statute, and two regulations. They are short, and they are quoted here in the words the courts have used, so that a reader can see for himself how little the law actually says about the practice.
The treaty
Article 3(1) of the Convention Against Torture and Other Cruel, Inhuman or Degrading Treatment or Punishment provides:
No State Party shall expel, return (“refouler”) or extradite a person to another State where there are substantial grounds for believing that he would be in danger of being subjected to torture.
One point about this text controls much of what follows. The D.C. Circuit has held that the Convention is not self-executing, and therefore “does not itself create any rights enforceable in U.S. courts,” citing the Supreme Court’s decision in Medellin v. Texas. Whatever the treaty requires of the United States as a matter of international law, a person cannot go into a federal court and sue on the treaty itself. He must find a statute.
The statute
That statute is section 2242 of the Foreign Affairs Reform and Restructuring Act of 1998, codified as a note to 8 U.S.C. section 1231. Its policy provision reads:
It shall be the policy of the United States not to expel, extradite, or otherwise effect the involuntary return of any person to a country in which there are substantial grounds for believing the person would be in danger of being subjected to torture, regardless of whether the person is physically present in the United States.
The reach of that sentence is wide. It covers expulsion, extradition, and any other means of effecting an involuntary return, and it applies whether or not the person is in the United States. But the Act also contains a review provision, and the D.C. Circuit has read the two together to a narrow result: “By its terms, the FARR Act provides a right to judicial review of conditions in the receiving country only in the immigration context, for aliens seeking review of a final order of removal.” On that reading, a person facing extradition or a military transfer has the benefit of the policy but no way to enforce it in court.
The regulations
The mechanism itself is regulatory. Section 1208.18(c) of title 8 of the Code of Federal Regulations, headed “Diplomatic assurances against torture obtained by the Secretary of State,” provides in full:
(1) The Secretary of State may forward to the Attorney General assurances that the Secretary has obtained from the government of a specific country that an alien would not be tortured there if the alien were removed to that country.
(2) If the Secretary of State forwards assurances described in paragraph (c)(1) of this section to the Attorney General for consideration by the Attorney General or her delegates under this paragraph, the Attorney General shall determine, in consultation with the Secretary of State, whether the assurances are sufficiently reliable to allow the alien’s removal to that country consistent with Article 3 of the Convention Against Torture. The Attorney General’s authority under this paragraph may be exercised by the Deputy Attorney General or by the Commissioner, Immigration and Naturalization Service, but may not be further delegated.
(3) Once assurances are provided under paragraph (c)(2) of this section, the alien’s claim for protection under the Convention Against Torture shall not be considered further by an immigration judge, the Board of Immigration Appeals, or an asylum officer.
The companion provision, section 1208.17(f), is headed “Termination pursuant to section 1208.18(c)” and provides that at any time while deferral of removal is in effect, the Attorney General may determine whether the deferral should be terminated on the basis of diplomatic assurances forwarded by the Secretary of State under those procedures.
What the regulations do not say
Three silences in that text are the whole of the litigation.
The question
What the regulation says
When may an assurance be sought and used?
Nothing. As the Third Circuit put it, section 1208.18 “provides no limitations on when diplomatic assurances may be invoked, either in terms of particular categories of aliens, or the status of an alien’s CAT claims in the adjudicatory process.”
What makes an assurance sufficiently reliable?
Nothing. The standard is stated but not defined, and no factors are listed.
What process is the person given?
Nothing. The Third Circuit found “nothing in the diplomatic assurance regulations themselves that we could fairly construe as providing an alien with any process whatsoever, let alone the right to a hearing.”
Paragraph (c)(3) is the sharpest of the three. Once the assurance is credited, the immigration judge, the Board, and the asylum officer are all removed from the matter by operation of the regulation. The finding of a probability of torture, which those adjudicators made, is displaced by a determination they play no part in and cannot revisit.
What the courts have done about that silence is the subject of the next section.
Four decisions carry most of the weight, and they do not point the same way. Read together they produce a rule that turns less on the danger the person faces than on the door he came through. An immigration case gets a hearing. An extradition case gets a signature. A military transfer gets neither.
Khouzam v. Attorney General (Third Circuit, 2008)
Sameh Sami S. Khouzam, an Egyptian citizen and a Coptic Christian, had been granted relief from removal after an adjudicator found that it was more likely than not that he would be tortured if returned to Egypt. His removal was deferred rather than withheld because there were serious reasons to believe he had committed a murder before leaving Egypt. He was released from custody in 2006.
In 2007 the Department of Homeland Security detained him again and prepared to remove him on the strength of assurances from Egypt that he would not be tortured. The court described what he was given, and the description is the case: “The Government provided no prior notice to Khouzam regarding the diplomatic assurances. Nor did the Government provide Khouzam any opportunity to review the assurances, or to present evidence or arguments challenging the assurances before an IJ, the BIA, or any other body.”
The Third Circuit held that this violated the Due Process Clause. “Because the Government violated the Due Process Clause by terminating Khouzam’s deferral of removal without affording him an opportunity to test the reliability of Egypt’s diplomatic assurances, the termination order was invalid.” The court then described what was required instead: “Prior to removal on the basis of diplomatic assurances, Khouzam must be afforded notice and an opportunity to test the reliability of those assurances in a hearing,” and “The alien must have an opportunity to present, before a neutral and impartial decision-maker, evidence and arguments challenging the reliability of diplomatic assurances proffered by the Government.” The court added that the person “must also be afforded an individualized determination of the matter based on a record disclosed to the alien.”
Two limits are worth noting, because they are often lost when the case is described. The court held that the district court had no jurisdiction over the habeas petition, and vacated the order granting it; the relief came through the petition for review instead. And the court did not hold that assurances from Egypt were categorically unreliable, which was Khouzam’s second argument. It held that he was entitled to make that argument somewhere.
Two American citizens held in Iraq by the Multi-National Force sought to block their transfer to Iraqi custody, arguing that transfer was likely to result in torture. The Supreme Court held that the federal courts had habeas jurisdiction but that the petitioners were not entitled to relief.
On the torture allegation the Court said: “Such allegations are of course a matter of serious concern, but in the present context that concern is to be addressed by the political branches, not the Judiciary.” It gave reasons of institutional competence, observing that the political branches “are well situated to consider sensitive foreign policy issues, such as whether there is a serious prospect of torture at the hands of an ally, and what to do about it if there is.”
The Court also recorded two things that limit the holding. It noted the Solicitor General’s statement that it is the policy of the United States not to transfer an individual in circumstances where torture is likely to result, and it expressly reserved a harder case: this, the Court said, “is not a more extreme case in which the Executive has determined that a detainee is likely to be tortured but decides to transfer him anyway.” Justice Souter, concurring, would have extended that reservation further, to “a case in which the probability of torture is well documented, even if the Executive fails to acknowledge it.”
Returning to the D.C. Circuit after Munaf, Omar argued that the Foreign Affairs Reform and Restructuring Act gave him a statutory right to judicial review of conditions in the receiving country. The court rejected the argument on the text: “By its terms, the FARR Act provides a right to judicial review of conditions in the receiving country only in the immigration context, for aliens seeking review of a final order of removal. The FARR Act does not give extradition or military transferees … a right to judicial review of conditions in the receiving country.”
The court acknowledged the asymmetry this creates and did not regard it as a defect: section 2242(a) “states a broad ‘policy’ that the Executive Branch presumably has a responsibility to follow with respect to all transfers,” while “only immigration transferees may obtain judicial review of conditions in the receiving country before they are transferred.” Congress, the court held, may create a right for one category of transferee without creating it for the others.
This is the decision that establishes the door rather than the danger as the operative fact. Two people facing the same risk in the same country have different rights depending on the legal process that brings them to the border.
Trinidad y Garcia v. Thomas (Ninth Circuit, en banc, 2012)
An extradition case, decided by a fractured en banc court in a per curiam opinion. The court held that it had jurisdiction, and that as a matter of due process the Secretary of State must consider the extraditee’s torture claim and must not extradite him if she finds it more likely than not that he will be tortured. It described the resulting entitlement narrowly: “An extraditee thus possesses a narrow liberty interest: that the Secretary comply with her statutory and regulatory obligations.”
The remedy is where the case ends up. Because the record contained no evidence that the Secretary had complied, the court remanded so that a declaration to that effect could be supplied. What happens then is set out in the opinion itself: “If the district court receives such a declaration, it shall determine whether it has been signed by the Secretary or a senior official properly designated by the Secretary. If so, the court’s inquiry shall have reached its end.” The court explained that the substantive claim “is foreclosed by Munaf v. Geren,” and that “the doctrine of separation of powers and the rule of non-inquiry block any inquiry into the substance of the Secretary’s declaration.”
The court divided sharply. Judge Berzon, joined by Judge W. Fletcher, concurred in the first five parts but declined to accept “that once the Secretary (or her delegate) meets the procedural due process requirement by submitting a barebones declaration, courts under no circumstances have authority to conduct any substantive review.” Judge Tallman, joined by three colleagues, would have gone the other way entirely, treating the rule of non-inquiry as a complete answer.
Notice, and a hearing before a neutral decision-maker at which the reliability of the assurance can be contested, on a record disclosed to the person
Khouzam (3d Cir. 2008)
Extradition
Confirmation that the Secretary of State signed a declaration of compliance. No inquiry into whether the underlying judgment was sound
Trinidad y Garcia (9th Cir. 2012) (en banc)
Military transfer abroad
No statutory right of review of conditions in the receiving country, and no freestanding constitutional right
Munaf (2008); Omar v. McHugh (D.C. Cir. 2011)
Two cautions about that table. It states the rule in the circuits that decided these cases, and a circuit that has not addressed the question is not bound by any of them. And it describes the position established by these four decisions, not the position in any later case; a reader relying on it for a live matter should confirm that each decision is still good law in the relevant circuit.
Decisions in the Coalition’s register
The four cases above are the doctrinal spine, and they are older than the removals now being reported. The Coalition’s own register holds the recent orders, read one by one and entered with what each of them decided. See the decisions tagged with assurances, and be aware that the tag is thin: assurances appear more often as a premise in these cases than as the question presented. The nearby categories are usually more productive.
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.
Everything asserted in this project comes from something on this page. Each item was opened and read at the address given. Where a source could not be opened, or did not state a detail, that is said rather than guessed at. Nothing is listed here that was not checked.
The primary law
Convention Against Torture and Other Cruel, Inhuman or Degrading Treatment or Punishment, article 3(1). The text used here is the text as quoted by the Supreme Court in Munaf v. Geren and by the D.C. Circuit in Omar v. McHugh.
Foreign Affairs Reform and Restructuring Act of 1998, section 2242, Pub. L. No. 105-277, 112 Stat. 2681, codified at 8 U.S.C. section 1231 note. Quoted at length in Omar v. McHugh.
8 C.F.R. section 1208.18(c) and 8 C.F.R. section 1208.17(f). Both reproduced in full in Khouzam v. Attorney General.
22 C.F.R. section 95.2, the State Department regulation applied in the extradition context, discussed in Trinidad y Garcia v. Thomas.
The decided cases
Khouzam v. Attorney General of the United States, 549 F.3d 235 (3d Cir. 2008), decided December 5, 2008, Nos. 07-2926 and 08-1094. Full text on CourtListener.
Omar v. McHugh, 646 F.3d 13 (D.C. Cir. 2011), decided June 21, 2011, No. 09-5410. Full text on CourtListener.
Trinidad y Garcia v. Thomas, 683 F.3d 952 (9th Cir. 2012) (en banc) (per curiam), decided June 8, 2012, No. 09-56999. Full text on CourtListener. CourtListener captions this record “Hedelito Garcia v. Linda Thomas”; the reporter caption is used on this site.
Three further decisions are relied on by the courts above and are named here so a reader can follow the chain: Kiyemba v. Obama (“Kiyemba II”), 561 F.3d 509 (D.C. Cir. 2009); Mironescu v. Costner, 480 F.3d 664 (4th Cir. 2007); and Cornejo-Barreto v. Seifert, 218 F.3d 1004 (9th Cir. 2000), which the Ninth Circuit overruled in part in Trinidad y Garcia. These are cited as the deciding courts cited them; the Coalition has not read them in full.
Scholarship arguing that assurances can be used
Ashley S. Deeks, “Avoiding Transfers to Torture,” Council on Foreign Relations, Council Special Report No. 35 (June 2008). Report page.
Ryan Goodman, “Forced Transfer of Detainees with Diplomatic Assurances Against Ill-Treatment,” Just Security (December 16, 2013). Read the post.
Aristi Volou, “Are Diplomatic Assurances Adequate Guarantees of Safety Against Torture and Ill-Treatment? The Pragmatic Approach of the Strasbourg Court,” UCL Journal of Law and Jurisprudence, vol. 4, issue 1 (2015). Article page. The repository copy of the PDF would not open; the article page was read instead.
Scholarship arguing that they cannot
Gregor Noll, “Diplomatic Assurances and the Silence of Human Rights Law,” Melbourne Journal of International Law, vol. 7 (2006). Full text, PDF.
Başak Çalı, Cathryn Costello and Stewart Cunningham, “Hard Protection through Soft Courts? Non-Refoulement before the United Nations Treaty Bodies,” German Law Journal, vol. 21, pp. 355 to 384 (2020). Full text, PDF.
Margit Ammer and Andrea Schuechner, “Article 3: Principle of Non-Refoulement,” in Nowak, Birk and Monina (eds), The United Nations Convention Against Torture and its Optional Protocol: A Commentary (2nd ed., Oxford University Press, 2019). Chapter page.
Brenna D. Nelinson, “From Boumediene to Garcia: The United States’ (Non)Compliance with the United Nations Convention Against Torture and its Movement Away From Meaningful Review,” American University International Law Review, vol. 29, issue 1, pp. 209 to 253 (2013). Full text, PDF.
On the safeguards a credible practice would need
Naureen Shah, Columbia Law School Human Rights Institute, “Promises to Keep: Diplomatic Assurances Against Torture in US Terrorism Transfers” (2010). Report page.
Steve Vladeck, “Diplomatic Assurances, Torture, and Judicial Review: The Bimenyimana Appeal,” Just Security (March 27, 2015). Read the post.
On the current removal program
Scott Roehm and Rebecca Ingber, “In Congress, a Welcome, and Well-Executed, Next Step to Stop Trump’s Transfers to Torture,” Just Security (August 8, 2025). Read the post.
Megan E. Elman-Welch, “Deportations to Perilous Places: Protecting Against Third Country Removal in Immigration Court,” Georgetown Immigration Law Journal, vol. 40, p. 383 (2026). Full text, PDF.
Matthew Boaz, “Due Process in Third Country Removals,” Lawfare (December 1, 2025). Read the article.
What is missing from this list
Two gaps are worth stating, because a reader is entitled to know the limits of what has been gathered. No law review case note on Khouzam was found, and none has been invented to fill the space. No article addressing 8 C.F.R. section 1208.18(c) on its own terms was found either; the regulation is treated here through the courts that have construed it and through the writers who discuss it in passing.
If you know of a source that belongs on this page, or an error in what is here, the Coalition wants to hear it. Corrections are made with the correction shown rather than silently.
Stay Informed
Receive updates on torture prevention, OPCAT ratifications, UN Special Rapporteur reports, and the Coalition campaigns worldwide.