FOR PUBLICATION

UNITED STATES COURT OF APPEALS

FOR THE NINTH CIRCUIT

DELFINA SOTO-SOTO, No. 20-70587 Petitioner, Agency No. v. A209-406-355

MERRICK B. GARLAND, Attorney General, OPINION Respondent.

On Petition for Review of an Order of the Board of Immigration Appeals

Argued and Submitted January 11, 2021 San Francisco, California

Filed June 11, 2021

Before: J. CLIFFORD WALLACE and MILAN D. SMITH, JR., Circuit Judges, and JANE A. RESTANI, * Judge.

Opinion by Judge Milan D. Smith, Jr.; Partial Concurrence and Partial Dissent by Judge Wallace

* The Honorable Jane A. Restani, Judge for the United States Court of International Trade, sitting by designation. 2 SOTO-SOTO V. GARLAND

SUMMARY **

Immigration

Granting Delfina Soto-Soto’s petition for review of a decision of the Board of Immigration Appeals’ reversing an immigration judge’s grant of deferral of removal under the Convention Against Torture, and remanding for the Board to grant CAT relief, the panel held that the Board erred by reviewing the IJ’s decision de novo, rather than for clear error, and concluded that the record compelled the conclusion that Soto-Soto met her burden of proof to establish that it is more likely than not that she will suffer future torture if removed to Mexico.

Michoacán state police arrested and brutally tortured Soto-Soto until she confessed to the kidnapping and murder of a five-year old boy. After the Mexican trial court dismissed the charges against her as a result of due process errors during the investigation, she fled to the United States. Mexican prosecutors subsequently conducted a new investigation and filed new charges against Soto-Soto, INTERPOL put out a Red Notice for her extradition to Uruapan in Michoacán, Mexico, which is 67 miles from where Soto-Soto was tortured in Morelia, Michoacán. Relying on Soto-Soto’s past torture, her reporting of the torture to the Michoacán State Commission of Human Rights despite warnings not to do so, the reissued arrest warrant, and country condition evidence showing that indigenous women like Soto-Soto are particularly

** This summary constitutes no part of the opinion of the court. It has been prepared by court staff for the convenience of the reader.

SOTO-SOTO V. GARLAND 3

vulnerable to torture, the IJ held that Soto-Soto was more likely than not to be tortured again if removed to Mexico. The Board reversed and held that the IJ’s determination was clearly erroneous because he did not acknowledge the Mexican judicial system’s appropriate steps to correct past due process errors, that Soto-Soto was not harmed while in custody for eight months after reporting the torture, and that members of Soto-Soto’s family remain in Mexico unharmed.

The panel concluded that the Board’s decision reflected that it engaged in a de novo weighing of the evidence, rather than clear error review. The panel explained that the Board may find an IJ’s factual finding to be clearly erroneous if it is illogical or implausible, or without support in inferences that may be drawn from the facts in the record, but in this case, the Board failed to explain how the IJ’s decision was illogical, implausible, or without support.

The panel’s majority also concluded that the Board’s view of the evidence was not supported by the record. First, the majority wrote that the record emphatically did not show that the Mexican court took steps to cure the due process errors caused by the state police officers torturing a confession out of Soto-Soto. Further, the majority wrote that even if the record supported the Board’s factual findings, that would not be enough to overturn the IJ’s decision under clear error review, because the IJ’s predictive finding as to the likelihood of torture is entitled to broad deference, which the Board failed to provide. Second, the majority wrote that because Soto-Soto’s human rights commission complaint was not filed until after she was released from custody, and nothing in the record suggested that the state police officers were aware of her report, Soto-Soto’s physical safety while in custody was not probative of the state police officers’ 4 SOTO-SOTO V. GARLAND

intent to carry out their threat of future torture. Finally, the majority wrote that the lack of harm to Soto-Soto’s family was irrelevant because threats of such harm hinged on Soto- Soto’s return to Mexico, which had not yet occurred. The panel also observed that the Board failed to discuss the IJ’s other key factual findings, including country condition reports establishing that indigenous women are more likely to be tortured in Mexico than other groups.

Reviewed under the proper standard of review, the majority concluded that the IJ’s decision was not clearly erroneous, and that the record compelled the conclusion that Soto-Soto met her burden of proof to establish that it is more likely than not that she will suffer future torture if removed to Mexico. The majority remanded the petition to the Board with the direction to grant deferral of removal.

Concurring in part and dissenting in part, Judge Wallace agreed that the Board impermissibly applied de novo review in reversing the IJ’s grant of relief. However, Judge Wallace wrote that he believes that the IJ erred in the likelihood of future torture analysis, and he relatedly disagreed with the majority’s and IJ’s conflation of the various Mexican law enforcement actors in the state of Michoacán into a unitary actor—i.e., the Michoacán state police—in assessing the likelihood of torture. Judge Wallace also highlighted that the IJ found Soto-Soto was ineligible for asylum and withholding of removal because there are serious reasons to believe that she did, in fact, kidnap and murder the child when his family refused to pay the demanded ransom. Judge Wallace believed that her likely guilt should have been considered as well because her original criminal case was not dismissed due to factual innocence but due process errors that have been corrected. Judge Wallace concluded that the majority’s direction to the Board to grant CAT relief rather

SOTO-SOTO V. GARLAND 5

than reversing and remanding the petition to the Board for further consideration goes against the court’s ordinary practice, especially because the record did not compel the conclusion that Soto-Soto satisfied her burden of proof.

COUNSEL

Hector A. Vega-Reyes (argued), Public Defender’s Office, San Francisco, California, for Petitioner.

Jeff Beelaert (argued) and Virginia Lum, Trial Attorneys; Justin Markel, Senior Litigation Counsel; Ethan P. Davis, Acting Assistant Attorney General; Office of Immigration Litigation, Civil Division, United States Department of Justice, Washington, D.C.; for Respondent.

OPINION

M. SMITH, Circuit Judge:

Petitioner Delfina Soto-Soto was brutally tortured by Mexican state police until she confessed to the kidnap and murder of five-year-old Bernardino Bravo Gomez. After the Mexican trial court dismissed the charges against her due to lack of evidence of the crimes charged, she fled to the United States. Mexican prosecutors subsequently refiled the charges against Soto-Soto, INTERPOL put out a Red Notice for her extradition to Mexico, and the Department of Homeland Security (DHS) placed her in removal proceedings. The Immigration Judge (IJ) granted deferral of removal under the Convention Against Torture (CAT) based on a factual finding that Soto-Soto was more likely than not to be tortured again if removed to Mexico. The Board of 6 SOTO-SOTO V. GARLAND

Immigration Appeals (BIA) reversed the grant of CAT relief and ordered Soto-Soto removed. Soto-Soto appealed. We vacate the BIA’s order and remand with instructions to grant relief.

FACTUAL AND PROCEDURAL BACKGROUND

Delfina Soto-Soto is a 42-year-old indigenous woman from Uruapan, a city in the Mexican state of Michoacán. On April 20, 2012, Michoacán state police—in plainclothes and driving unmarked vehicles—arrested Soto-Soto at her home in Uruapan for allegedly kidnapping and murdering Bernardino. The police drove her to Morelia, a town approximately two hours away from Uruapan. She initially denied any involvement in the kidnapping.

Michoacán state police brutalized her for hours. According to Soto-Soto’s testimony, the police first grabbed her by her hair and threw her to the ground. While she was on the ground, they tied her hands behind her back, placed her face-up, and kicked her repeatedly. One officer held her head between his knees while the others poured water into her nose and punched her stomach. The officers then put a plastic bag over her head and sat on her stomach so that she was unable to breathe, all the while calling her a “fucking bitch.”

The torture continued to escalate as Soto-Soto refused to admit involvement in the kidnapping. The police officers slapped her face, threw her to the ground, tied the plastic bag over her head until she was at the point of suffocation, and then repeated the sequence for hours. Next, the officers sat her in a chair and hit her head against the wall. One officer pointed a gun at her head several times, telling her that if she did not sign a confession, they would “keep [her] right there like a bitch.” Still, with the gun pressed against her forehead,

SOTO-SOTO V. GARLAND 7

Soto-Soto did not relent. Finally, as relayed in Soto-Soto’s testimony, one of the officers told Soto-Soto, “[I]f you don’t sign this [confession] in the next 15 minutes, I will give the order for them to go get your daughters and bring them here, first they will rape them and then they will kill them in front of you.” At that point, Soto-Soto signed the confession. The officers also told her that she would be tortured again if she reported their acts to anyone.

Soto-Soto was held in custody in Uruapan for the next eight months and was not tortured again. On December 10, 2012, the Mexican court dismissed the charges for lack of evidence and released Soto-Soto from custody. Upon her release, Soto-Soto immediately fled to the United States. On December 19, 2012, the Michoacán State Human Rights Commission filed a complaint about the torture on Soto- Soto’s behalf, but it was dismissed two days later because Soto-Soto did not provide an address where she could be contacted.

In 2013, the Government renewed its investigation of Soto-Soto and secured an arrest warrant after presenting testimony of sixteen witnesses to the Mexican trial court. Soto-Soto contends that these witnesses’ testimony also underlay the first prosecution, and that the witnesses are individuals from her village who believe that she is a witch. On November 12, 2015, INTERPOL issued a Red Notice for Soto-Soto, which called for her return to the state of Michoacán. She was taken into custody from her workplace in Madera, CA on November 28, 2017.

After a multi-day hearing, the IJ found that Soto-Soto was credible and accorded her testimony full evidentiary weight. The IJ further held that Soto-Soto was statutorily ineligible for asylum, withholding of removal under the INA, and withholding of removal under the CAT because 8 SOTO-SOTO V. GARLAND

the warrant for her arrest in the second investigation provided “serious reasons to believe” that she committed a “serious nonpolitical crime” before arriving in the United States.

The IJ also considered whether Soto-Soto was eligible for deferral of removal under CAT, concluding that the alleged serious nonpolitical crime did not bar her from this remedy. In order to obtain deferral of removal, Soto-Soto had the burden to show that it was more likely than not that she would be tortured with government involvement or acquiescence if removed to Mexico. 8 C.F.R. §§ 1208.16(c)(2), 1208.17(a).

In the IJ’s evaluation of the likelihood of future torture, the IJ first found that Soto-Soto had suffered torture at the hands of Michoacán state police from the Office of Anti- Kidnapping and Extortions. Based on the reissued arrest warrant in the second prosecution, the IJ further found that it was more likely than not that Soto-Soto would be stopped by authorities upon her return and taken back into custody. The IJ noted that country condition evidence for Mexico documents that indigenous women like Soto-Soto are particularly vulnerable to torture. Finally, the IJ determined that the fact that Soto-Soto reported the previous torture to the Michoacán State Commission of Human Rights also makes it more likely that she will be tortured—the state police officers specifically threatened her with this. Based on these facts, the IJ found that Soto-Soto had satisfied her burden to show that she would likely suffer torture with government involvement or acquiescence if removed to Mexico. The Government appealed.

The BIA reversed the IJ’s grant of deferral of removal under CAT. The BIA held that the IJ’s findings were clearly erroneous because he did not acknowledge the Mexican

SOTO-SOTO V. GARLAND 9

judicial system’s appropriate steps to correct past due process errors, that Soto-Soto was not harmed while in custody for 8 months after reporting the torture, and that members of Soto-Soto’s family remain in Mexico unharmed.

With respect to the first alleged error, the BIA further wrote, “[T]he Immigration Judge clearly erred in finding that the Mexican authorities ‘reinitiated the prosecution against the Respondent,’ without acknowledging that the Mexican judicial system excluded the evidence obtained through torture and conducted a new investigation that independently resulted in the current charges against the respondent.” Furthermore, the BIA stated, “The Immigration Judge clearly erred in finding that the respondent’s HRC complaint was dismissed on a technicality (IJ at 9); the respondent testified that she did not attempt to pursue the claim or contact the HRC after she was released and departed the country.” 1 For these reasons, the BIA held that the IJ committed clear error and reversed the IJ’s grant of CAT deferral of removal. Soto-Soto appealed.

On appeal, Soto-Soto contends that—though the BIA’s opinion says it reviews the IJ’s finding under the clear error standard—the substance of the BIA opinion improperly engages in de novo review. We agree.

1 It is undisputed that HRC did not file Soto-Soto’s complaint of torture until after she was released from custody, meaning that while Soto-Soto was in custody, the state police were not aware that she had disclosed the torture. 10 SOTO-SOTO V. GARLAND

STANDARD OF REVIEW

“Whether the BIA has applied the correct standard of review is a question of law” that we review de novo. Rodriguez v. Holder, 683 F.3d 1164, 1169 (9th Cir. 2012).

ANALYSIS

A.

The parties agree that the BIA ought to have reviewed the IJ’s factual findings for clear error. As this court has stated:

Where the BIA engages in de novo review of an IJ’s factual findings instead of limiting its review to clear error, it has committed an error of law, as our sister circuits have recognized, and we have no difficulty in agreeing with that conclusion. We do not rely on the Board’s invocation of the clear error standard; rather, when the issue is raised, our task is to determine whether the BIA faithfully employed the clear error standard or engaged in improper de novo review of the IJ’s factual findings.

Rodriguez v. Holder, 683 F.3d 1164, 1170 (9th Cir. 2012) (citations and footnote omitted).

To determine whether “the BIA faithfully employed the clear error standard,” courts look at the BIA’s reasoning for reversing the IJ’s decision. Id. “The BIA may find an IJ’s factual finding to be clearly erroneous if it is illogical or implausible, or without support in inferences that may be drawn from the facts in the record.” Id. at 1170 (internal

SOTO-SOTO V. GARLAND 11

quotation marks omitted). Where the BIA does not address the “key factual findings” on which the IJ based its conclusion, that strongly suggests that the BIA did not faithfully engage in clear error review. Zumel v. Lynch, 803 F.3d 463, 475, 476 (9th Cir. 2015).

Furthermore, if it appears that the BIA gave more weight to certain facts in the record than to others, leading to a different conclusion from the IJ, our court may justifiably infer that the BIA applied the wrong standard of review. See Guerra v. Barr, 974 F.3d 909, 914 (9th Cir. 2020). In Guerra, we reversed the BIA for reviewing the IJ’s decision de novo, rather than for clear error, reasoning that “the clear error standard does not allow the BIA to reweigh the evidence when the IJ’s account of the evidence is plausible.” Id. (citing Rodriguez, 683 F.3d at 1171). In explaining our reasoning, we wrote:

The BIA assumed without deciding that Guerra faces a likelihood of being institutionalized in a Mexican mental health institution. But it rejected the IJ’s finding of specific intent, noting “there is insufficient record evidence from which it is reasonable to conclude that health care workers implement such [primitive and abusive] practices for the specific purpose of inflicting harm on the patients.” Instead, the BIA accorded more weight to country reports in the record that the extreme measures were taken as a misguided effort to prevent patients from harming themselves or others.

Id. at 913–14 (alterations in original). In other words, “[a]n appellate court cannot substitute its interpretation of the 12 SOTO-SOTO V. GARLAND

evidence for that of the trial court simply because the reviewing court might give the facts another construction [or] resolve the ambiguities differently[.]” Inwood Labs., Inc. v. Ives Labs., Inc., 456 U.S. 844, 857–58 (1982) (internal quotation marks omitted).

In Soto-Soto’s case, the IJ found: (1) Soto-Soto had suffered torture in the past; (2) she would likely be arrested if returned to Mexico, putting her back within reach of her torturers; (3) her status as an indigenous woman made her more likely to be tortured according to the country reports; and (4) her report of the torture to the human rights commission increased her likelihood of torture.

The BIA disagreed with the IJ’s view of the evidence. But its only explanation of why the IJ’s decision was illogical, implausible, or without support was that “[the IJ] did not acknowledge that ‘the Mexican judicial system took appropriate steps to correct any past due process errors committed by the officers of the Office of Anti-kidnapping and Extortions,’ that the respondent reported the torture and was not subsequently harmed or threatened while in custody for nearly 8 months, and that other members of her family have remained unharmed in Mexico.” The BIA did not explain how these alleged errors showed lack of logic, plausibility, or support in the record on the part of the IJ. The BIA’s reasoning is therefore insufficient to demonstrate that the BIA engaged in clear error review. Instead, the BIA’s reasoning reflects a de novo weighing of the evidence.

Moreover, the BIA’s view of the evidence lacks support in the record. First, the record emphatically does not show that the Mexican court took steps to cure the “due process

SOTO-SOTO V. GARLAND 13

errors” 2 caused by the state police officers torturing a confession out of Soto-Soto. Instead, the record shows that the Mexican court issued “a court order dismissing charges . . . for lack of evidence in accordance with the law.” As part of its conclusion that the evidence against Soto-Soto was insufficient, the Mexican court held that Soto-Soto’s confession lacked probative value because “police carried out questionings at various domiciles without judicial authorization, nor permission from the owners,” and because there were “several contradictions regarding . . . the confession of the accused[.]” Additionally, the court determined that Soto-Soto’s confession lacked probative value because she did not have a Purepecha interpreter during the questioning. Notably absent from this order is any acknowledgment that state police had tortured Soto-Soto to obtain her confession, a bar upon her re-prosecution, or punishment for the officers who tortured her. The court also did not impose safeguards to ensure that the reopened investigation would not result in further torture or another improper confession. The record therefore contains no facts to support the BIA’s finding that the due process errors in Soto-Soto’s case were remedied.

Furthermore, even if the record supported the BIA’s factual findings, that would not be enough for the BIA to overturn the IJ under the clear error standard of review. Inwood Labs., Inc., 456 U.S. at 857–58. The IJ’s predictive fact finding leading to the holding that Soto-Soto was likely to be tortured again upon return to Mexico recognized that the second investigation resulted in the issuance of a warrant for her arrest. But in its discretion as factfinder, the IJ reasoned that this increased her likelihood of torture—rather

2 Though it strains credulity to call the brutal torture Soto-Soto endured a “due process error,” we use the BIA’s terms for consistency. 14 SOTO-SOTO V. GARLAND

than decreasing it—because it meant that Soto-Soto would almost certainly end up in the hands of the state police who had previously tortured her and had threatened to do so again. This type of fact finding—how the facts in the record affect the likelihood of future torture—is entitled to broad deference from the BIA. See Vitug v. Holder, 723 F.3d 1056, 1063 (9th Cir. 2013). The BIA’s review of that predictive fact finding, however, exhibits no deference to the IJ’s decision.

The second reason the BIA gave for reversing the IJ— that Soto-Soto was unharmed for eight months after disclosing the prior torture—is also unmoored from the record. The human rights commission complaint was not filed until after Soto-Soto was released from custody. Nothing in the record suggests that the state police officers, who threatened to torture her if she reported the abuse, were aware that she had done so. Soto-Soto’s physical safety while in custody is therefore not probative of the state police officers’ intent to carry out their threat of future torture if Soto-Soto reported their past torture.

The BIA’s final point, that Soto-Soto’s family members remain unharmed in Mexico, is irrelevant. The current threats to Soto-Soto’s family members hinge on whether Soto-Soto returns to Mexico. For example, Soto-Soto’s daughter, Leonarda Bravo Soto, wrote, “I still receive threats that if my mother returns[,] . . . they will take my daughter away. . . . [We], her children, do not want her to return to the town. . . . We love our mother so much and want her alive for many years more.” Because Soto-Soto has not returned to Mexico, the current physical safety of her family members does not indicate that the threats against them are baseless.

SOTO-SOTO V. GARLAND 15

Moreover, the BIA failed to discuss the IJ’s other “key factual findings,” specifically that country condition reports established that indigenous women are more likely to be tortured in Mexico than other groups. See Zumel, 803 F.3d at 475, 476. The BIA stated only that “the Mexican government has made strident efforts to combat the use of torture[.]” In concluding that the government’s “strident efforts” to eliminate torture outweighed the country conditions report showing that indigenous women are at an increased risk of torture, the BIA re-weighed evidence and failed to engage in clear error review. Similarly, in countering the IJ’s finding that community members had credibly threatened Soto-Soto, the BIA wrote that when these individuals were actively beating Soto-Soto’s co- defendant, “the local police did arrive and intervene to some degree[.]” But the local police who arrived testified that they did not stop the crowd from beating the co-defendant. The BIA’s assertion is, again, both unfounded and reflective of an independent weighing of the evidence rather than clear error review.

Based on this analysis, the BIA’s decision reflects de novo, rather than clear error, review. The BIA therefore applied the wrong standard of review, substituting its own view of the evidence for the IJ’s.

B.

Reviewed under the proper standard of review, the IJ’s factual findings were not clearly erroneous. A court “assessing whether it is more likely than not that an applicant would be tortured in the proposed country of removal” must consider:

(i) Evidence of past torture inflicted upon the applicant; 16 SOTO-SOTO V. GARLAND

(ii) Evidence that the applicant could relocate to a part of the country of removal where he or she is not likely to be tortured;

(iii) Evidence of gross, flagrant or mass violations of human rights within the country of removal, where applicable; and

(iv) Other relevant information regarding conditions in the country of removal.

8 C.F.R. § 1208.16(c)(3). Of these,

[p]ast torture is ordinarily the principal factor on which we rely when an applicant who has been previously tortured seeks relief under [CAT] because, absent changed circumstances, if an individual has been tortured and has escaped to another country, it is likely that she will be tortured again if returned to the site of her prior suffering.

Xochihua-Jaimes v. Barr, 962 F.3d 1175, 1188 (9th Cir. 2020) (cleaned up).

In accordance with the items listed in 8 C.F.R. § 1208.16(c)(3), the IJ found that the Michoacán state police tortured Soto-Soto, and the revived arrest warrant guaranteed she would be placed back in custody of the Michoacán state police, who previously tortured her,

SOTO-SOTO V. GARLAND 17

precluding relocation. 3 Moreover, the state police officers specifically threatened to torture Soto-Soto again if she reported their misconduct—which she did. 4 Finally, the IJ

3 Our dissenting colleague believes that we improperly conflate “the various Mexican law enforcement actors in the state of Michoacán into a unitary actor,” because there is no evidence that the Michoacán Office of Anti-Kidnapping and Extortions in the city of Morelia is a “component” of the Michoacán police force in the city of Uruapan, where Soto-Soto would have been held for trial. We disagree that the record lacks such evidence. For instance, the Human Rights Commission Complaint filed on Soto-Soto’s behalf states that she was tortured by “members of the State Ministry Police of the Office of Anti- kidnapping and Extortions of the State Attorney General.”

Furthermore, the Anti-Kidnapping Office found Soto-Soto at her home in Uruapan and took her to Morelia to brutalize her before she was transferred back to Uruapan to be held in prison pending trial. Uruapan and Morelia are both in the state of Michoacán. It would be illogical to conclude that, although Soto-Soto was not safe from the “State Ministry Police of the Office of Anti-kidnapping and Extortions” in her home in Uruapan, she would be safe from them in a jail cell there, watched over by other state police. Moreover, the BIA never concluded—and the Government never argued—that Soto-Soto was ineligible for CAT relief because the Michoacán Anti-Kidnapping Office was separate from the Michoacán state police. We therefore respectfully disagree with Judge Wallace that it would be appropriate for us to base our decision on this waived factual issue. See Clem v. Lomeli, 566 F.3d 1177, 1182 (9th Cir. 2009) (stating that, in general, appellees waive issues that they fail to raise in their answering brief); United States v. Dreyer, 804 F.3d 1266, 1277–78 (9th Cir. 2015) (en banc) (same); Moran v. Screening Pros, LLC, 943 F.3d 1175, 1180 (9th Cir. 2019) (same). 4 Our dissenting colleague need not be concerned that our disposition of Soto-Soto’s case improperly relies on the dissent in Dawson v. Garland, Case No. 19-73142, 2021 WL 2125800 (9th Cir. May 26, 2021). Dissent at 23. In Dawson, the majority affirmed the denial of CAT relief where the petitioner suffered past torture at the hands of her intimate partner before obtaining a protective order that stopped the torture because “[t]he inference that future torture is likely 18 SOTO-SOTO V. GARLAND

considered the country condition reports showing an increased threat of torture for indigenous women. The IJ did not err, and a proper application of the clear error standard of review would have resulted in the BIA affirming the IJ’s grant of relief. Indeed, the record compels the conclusion that Soto-Soto “met her burden of proof to establish that it is more likely than not that she will suffer future torture if removed to her native country.” Id. at 1188.

CONCLUSION

Because the record compels the conclusion that Soto- Soto carried her evidentiary burden, we GRANT the petition and REMAND for the BIA to grant deferral of removal pursuant to CAT. Id. (citing Haile v. Holder, 658 F.3d 1122, 1133 (9th Cir. 2011)).

to recur breaks down where ‘circumstances or conditions have changed significantly, not just in general, but with respect to the particular individual.’” 2021 WL 2125800, at *6 (quoting Nuru v. Gonzales, 404 F.3d 1207, 1218 (9th Cir. 2005)). Dawson does not apply to this case because—as set forth above—Soto-Soto’s circumstances have not significantly changed.

Moreover, the rule that past torture is the principal factor for evaluating the likelihood of future torture was established long before Dawson. See Xochihua-Jaimes, 962 F.3d at 1188; Avendano-Hernandez v. Lynch, 800 F.3d 1072, 1080 (9th Cir. 2015); Nuru, 404 F.3d at 1217– 18. By relying on that rule here, we tread no new or uncertain ground.

SOTO-SOTO V. GARLAND 19

WALLACE, Senior Circuit Judge, concurring in part and dissenting in part:

I concur with the majority’s holding limited to the issue that the Board of Immigration Appeals (Board) impermissibly applied de novo review when it reversed the Immigration Judge’s (IJ) grant of relief pursuant to the Convention Against Torture (CAT) to Soto-Soto. The Board’s use of “clear error” before each respective conclusion does not automatically mean it conducted clear error review. See Vitug v. Holder, 723 F.3d 1056, 1063 (9th Cir. 2013). The Board, instead, reweighed the evidence here when it emphasized factors ignored by the IJ rather than rejecting reasons relied upon by the IJ as illogical or impermissible inferences. See Rodriguez v. Holder, 683 F.3d 1164, 1170–71 (9th Cir. 2012) (acknowledging the Department of Justice’s position that “[a] factfinding may not be overturned simply because the Board would have weighed the evidence differently or decided the facts differently had it been the factfinder”) (alteration in original) (citations omitted). This is further evinced by the fact that the Board did not discuss the role of Soto-Soto’s status as an indigenous woman in its analysis. See Vitug, 723 F.3d at 1064 (holding that the Board “abuses its discretion where it ignores arguments or evidence”). At the most basic level, the Board should have conducted a more thorough analysis to explain how the IJ’s decision was illogical or not based upon permissible inferences from the record to comply with the clear error review standard. See Anderson v. Bessemer City, 470 U.S. 564, 577 (1985).

Had we gone no further and remanded to the Board for reconsideration, I could have joined in the opinion. However, the majority decided to add more, resulting in my dissent on two grounds. First, while I agree that the Board 20 SOTO-SOTO V. GARLAND

conducted impermissible de novo review, I nonetheless believe the IJ erred in his likelihood of future torture analysis when he granted CAT relief. Relatedly, I disagree with the majority’s conflation of the various Mexican law enforcement actors in the state of Michoacán into a unitary actor—i.e., the “Michoacán state police.” See, e.g., Maj. Op. at 6. In my opinion, the majority never refers to Soto-Soto’s torturers properly because it refers to them as the “Michoacán state police,” even though the more accurate description, which is used by the Board, is that they were members of the Office of Anti-Kidnapping and Extortions in Morelia, Michoacán. There is no evidence in the record to support the conclusion that the officers from the Office of Anti-Kidnapping and Extortions in Morelia are a component of the same Michoacán state police force that is in the city of Uruapan, where Soto-Soto would have been held for her trial. The two cities are 67 miles apart and there is no evidence of connection, legal or otherwise, of the police units of the two cities.

In my view, the IJ committed the same error during his likelihood of future torture analysis when he assumed that Soto-Soto would be returned to her torturers stationed at the Office of Anti-Kidnapping and Extortions in Morelia, Michoacán, despite INTERPOL’s specific direction in the Red Notice to return Soto-Soto to the court in Uruapan, Michoacán, where she, in fact, safely remained in custody for eight months waiting to be tried for the kidnaping and murder of a child from her neighborhood when the child’s family did not pay the demanded ransom.

Second, I dissent from the majority’s direction to the Board to grant CAT relief rather than reversing and remanding the petition to the Board for further consideration as is our ordinary practice. See Guerra v. Barr, 974 F.3d

SOTO-SOTO V. GARLAND 21

909, 917 (9th Cir. 2020) (observing that, where the Board “failed to apply the proper standard of review, we have generally vacated the agency’s decision and remanded for the [Board] to apply the appropriate standard of review”); Ornelas-Chavez v. Gonzalez, 458 F.3d 1052, 1058 (9th Cir. 2006) (reversing and remanding because the Board applied the incorrect legal standard).

I.

Even if the issue were properly before us, I would disagree with the majority’s decision to grant Soto-Soto’s request for CAT relief because I do not believe the evidence compels such a conclusion. From my reading of the record, the IJ erred in granting CAT relief due to two assumptions he made during the likelihood of future torture analysis that are neither supported by the record nor qualify as permissible inferences. First, the IJ assumed that Soto-Soto would be returned to the custody of the officers who tortured her in Morelia, Michoacán, despite explicit instructions in the Red Notice that Soto-Soto be transferred to authorities in Uruapan, Michoacán. As stated earlier, the two cities are approximately 67 miles apart. The IJ’s assumption was implausible because he ignored that Soto-Soto would have been prosecuted in Uruapan, Michoacán, for the new criminal charges against her so that she would have been in the custody of a separate law enforcement group and under the authority of a separate court. Uruapan is a separate jurisdiction within the state of Michoacán where, in fact, she remained unharmed for eight months awaiting trial, thereby proving the implausibility of the IJ’s assumption.

Second, the IJ merged the various law enforcement actors involved into the unitary concept of “Mexican authorities” rather than the various discrete law enforcement authorities throughout the state of Michoacán, i.e., the Office 22 SOTO-SOTO V. GARLAND

of Anti-Kidnapping and Extortions from Morelia, the Michoacán state police in Uruapan, and the judiciary in Uruapan. This non-record based unitary law enforcement theory muddies the actual factual situation. The IJ made no findings that would support such generalizations. We do not have the authority to find facts on appeal, and the factual issue should be resolved on remand. Cf. Ornelas-Chavez, 458 F.3d at 1060 (acknowledging that we cannot make factual findings and remanding the petition because we should not conduct an “independent review of the evidence” to determine whether a petitioner is eligible for CAT relief when applying the correct legal standard in the first instance). The majority repeats this error with its inclusive use of “Michoacán state police.”

The majority asserts that this factual issue has been waived because it was not raised by the parties; nonetheless, an impermissible inference was made by the IJ. Although the government did not make this explicit argument, it was careful to distinguish between the different law enforcement actors throughout the proceedings before the Agency and in its answering brief. The Board also acknowledged the factual distinction in its decision when it observed that Soto- Soto’s torturers were from the Office of Anti-Kidnapping and Extortions, as well as highlighted the torture in Morelia as opposed to Uruapan.

In addition, the majority’s citation to our decision in Clem v. Lomeli, 566 F.3d 1177 (9th Cir. 2009), to support its point that this factual distinction issue was waived is unavailing because Clem is readily distinguishable from this appeal. Clem was a Section 1983 prisoner case, in which Clem appealed from an adverse jury verdict and argued that the district court gave erroneous jury instructions and that the error was not harmless. Clem, 566 F.3d at 1179. We

SOTO-SOTO V. GARLAND 23

held in Clem that the jury instructions were erroneous. Id. at 1181–82. We next reasoned that the defendant could not overcome the presumption of prejudice because he failed to address prejudice in his answering brief. Id. at 1182. The majority relies on this point to argue that I should not consider the law enforcement distinction point when determining that the IJ made an impermissible inference. However, the majority forgets that the standard of review is different here, and we must consider whether substantial evidence supports the IJ’s determination that Soto-Soto is entitled to CAT relief in determining whether to remand her petition for reconsideration or to grant her request for CAT relief. Ultimately, I cannot ignore the IJ’s impermissible inference nor the majority’s own factual assumptions to reach its conclusion.

Finally, I observe that many of the arguments and assumptions the majority makes to support its conclusion that Soto-Soto is entitled to CAT relief were rejected by our court in a separate, unrelated opinion. See Dawson v. Garland, Case No. 19-73124, 2021 WL 2125800 (9th Cir. May 26, 2021) (Circuit Judges M. Smith and Ikuta, and District Judge Vratil). As I observe in my partial dissent, the majority in Dawson reasoned that the petitioner’s individual circumstances had changed and the lack of general changes in the country should not detract from her personally changed circumstances. Id. at *8. The Dawson dissent contended that petitioner’s past torture was the “principal factor” for evaluating the likelihood of future torture. Id. The majority makes a substantially similar argument here, with which I disagree.

As to the first point, Soto-Soto carries the burden to establish that it is more likely than not that she would be tortured if returned to Mexico to be granted CAT relief, in 24 SOTO-SOTO V. GARLAND

this case deferral of removal. 8 C.F.R. §§ 1208.16(c), 1208.17(a). Her application should have been reviewed under the standards for eligibility set out in section 1208.16(c)(3). See § 1208.17(a); see also Hamoui v. Ashcroft, 389 F.3d 821, 827 (9th Cir. 2004) (holding that petitioner must prove there is a “chance greater than fifty percent” that they “will be tortured if removed”). While “past torture is ordinarily the principal factor on which we rely when an applicant who has been previously tortured seeks” CAT relief, Xochihua-Jaimes v. Barr, 962 F.3d 1175, 1188 (9th Cir. 2020) (cleaned up), we must also consider whether the circumstances have changed.

Soto-Soto has established her past torture at the hands of the Office of Anti-Kidnapping and Extortions in Morelia, but there is no evidence in the record that supports imputing its conduct to the Michoacán state police in Uruapan, or that the separate courts in Uruapan would tolerate such activity. Yet the IJ assumed without record citation that the Michoacán state police in Uruapan are the same as the Office of Anti-Kidnapping and Extortions in Morelia. This was not a permissible inference based on the evidence in the record. Moreover, I cannot agree with the IJ’s and the majority’s refusal to consider the evidence of changed circumstances here that mitigate against the likelihood of future torture. I believe the Red Notice’s instruction to return Soto-Soto to Uruapan is a changed circumstance that shifts the balance against Soto-Soto’s claim. She was detained eight months in Uruapan, and there is no evidence that any member of the Michoacán state police in Uruapan harmed Soto-Soto or threatened to harm her. There is also no evidence that the Michoacán state police in Uruapan, or the Uruapan court, would permit Soto-Soto’s torturers from the Office of Anti- Kidnapping and Extortions in Morelia to torture her again before her trial or that the police in Uruapan would torture

SOTO-SOTO V. GARLAND 25

Soto-Soto as retribution for reporting her torture by the Office of Anti-Kidnapping and Extortions.

Even though Soto-Soto argues that the new criminal charges against her are based on the same evidence but without her coerced confession, so that she will likely be tortured again for a new confession, this argument is pure speculation and without support in the record. The argument also obscures the fact that the new investigation into the child’s kidnapping and murder still found sufficient evidence to once again charge Soto-Soto for the crimes, even without her coerced confession. The IJ also held that the serious, non-political crime bar applied because there are serious reasons to believe that Soto-Soto kidnapped and murdered the child. It is, therefore, important to emphasize that Soto-Soto’s original criminal case was not dismissed due to her torture or her factual innocence, but other due process errors that the court in Uruapan has corrected.

As to my second point, the record does not support the IJ’s assumption that the various police and judiciary actors in Michoacán, Mexico are a unitary actor of “Mexican authorities”, nor does it support the majority’s similar merger of authorities as the “Michoacán state police.” We have emphasized the importance of distinguishing between local, state, and federal police or countrywide efforts in persecution cases. See, e.g., Madrigal v. Holder, 716 F.3d 499, 510 (9th Cir. 2013). While the circumstances presented here are different, I believe the ultimate point rings true that we are required to consider which law enforcement group was responsible for the torture and whether that group would be able to torture her again if she were returned to Mexico.

Yet the majority fails to do so here. My concern is that the majority’s holding will lead to the perverse result that any victim of torture by law enforcement from one city can 26 SOTO-SOTO V. GARLAND

essentially claim an omnipresent threat from law enforcement throughout that region of the country, regardless of individual, ameliorative efforts taken by other law enforcement branches in the region. This is particularly troubling due to the lack of evidence in this case to support such a theory of CAT relief. Indeed, Soto-Soto was ordered to be returned to a separate court, 67 miles away from where the Office of Anti-Kidnapping and Extortions acted, to be tried for the kidnaping and murder of a neighbor’s child. There is no evidence that that court in Uruapan would allow Soto-Soto’s torture.

II.

Finally, I believe the majority errs by directing the Board to grant Soto-Soto CAT relief outright rather than remanding her petition to the Board to consider our opinion. Our ordinary practice is to remand a petition when the Board applied the incorrect standard, as it did here, unless there is a special circumstance that requires us to grant the underlying relief. See Guerra, 974 F.3d at 916–17. As I do not agree with the majority’s conclusion that the evidence compels us to grant Soto-Soto CAT relief, I also do not believe remanding without requiring further consideration by the Board is appropriate here. We should not interfere with our usual role as an appellate court to decide factual issues. Our usual remand to the Board procedure would overcome this mistake.

III.

For these reasons, I concur with the majority’s holding that the Board improperly engaged in de novo review and dissent from the majority’s failure to remand to the Board for reconsideration to determine whether Soto-Soto is entitled to CAT relief.