Justice Department Inspector General Michael Horowitz’s long‐​awaited report on “Crossfire Hurricane”—the FBI’s investigation of potential links between Russian election interference and the Trump campaign—has finally been released.

Like most news developments in our polarized age, the report is being spun in diametrically opposed ways by political partisans, as evidenced by the questions at Wednesday’s Senate Judiciary Committee hearing on the report. Both of these narratives, unfortunately, get it wrong in fundamental ways.

For many Democrats and other Trump critics, the main takeaway from the Horowitz Report has been that, despite some “irregularities” or “missteps” by the FBI, the report doesn’t support Trump or his allies’ claims about a “Deep State Coup” that ginned up a phony investigation to damage the Trump administration. Former FBI Director James Comey even treated the report as a vindication of the Bureau’s conduct during his tenure.

For many Republicans, the fact that some of the officials involved in a grossly flawed investigation of former Trump advisor Carter Page also had strongly negative views of Trump is all that’s needed to validate the view that the myriad serious failures Horowitz documents show “political abuse” motivated by partisan bias and a desire to “get Trump” at any cost. Both of these reads, I think, fundamentally misunderstand the real problems Horowitz exposed.

I’ve already done a lengthy analysis of the report’s central findings in a post over at Just Security, which I won’t try to exhaustively replicate here, but here are the core problems with these two accounts.

The “vindication” narrative makes this story—like every other story these days, it seems—essentially Trump‐​centric, so that the most important fact about the report is whether it agrees with or contradicts various outlandish things Trump has said.

If that’s your primary concern, then that narrative is mostly correct within its myopic domain: The report does not bear out, and in crucial respects is at odds with, Trump’s conspiratorial account of the investigation as part of some elaborate coup plot.

But as Charlie Savage rightly notes, this focus sidelines and trivializes the central findings of the report seen on its own terms: An appalling, systematic series of failures by the FBI that led to intrusive surveillance of an American citizen for nearly a year, when it should have been amply been clear by the time the first of three renewal applications was submitted that the original grounds for that surveillance were profoundly flawed.

The “political bias” narrative, though, ultimately makes an inverted version of the same error, insisting on seeing the facts Horowitz documents primarily (or exclusively) through the lens of partisan conflict: Pro‐​Trump, or anti‐​Trump? While it’s certainly possible that hostility to Trump played some role in misjudgments by individual officials, this doesn’t fit the facts on the whole well either.

It may be part of the story, but it is unlikely to be the most important part. This misdiagnosis matters, because if you think the problems Horowitz identifies are the function of “a few bad actors” misusing the process for reasons specific to the politically sensitive nature of the investigation, then however egregious the problems documented in the report, they don’t add up to a case for systemic reform:

You just need to fire (and maybe even prosecute) the specific bad actors, and perhaps add some more robust safeguards for the very tiny fraction of investigations that touch on domestic politics.

Here, I think is a more accurate account of what the report shows us. In the summer of 2016, against the backdrop of an elaborate and multi‐​pronged campaign of election interference by the Russian government, the FBI gets (accurate) intelligence from an allied government, Australia’s, suggesting that a Russian cutout has approached at least one Trump campaign staffer with what sounds like an offer of assistance.

They appropriately open a broad investigation into whether there are more extensive ties between the campaign and the Russian election operation. The probe quickly focuses on four campaign officials with links to Russia, among them foreign policy advisor Carter Page, who is already the subject of a counterintelligence investigation by the FBI’s New York field office, for reasons unrelated to the presidential campaign.

FISA surveillance on two of the four is contemplated, but initially rejected for lack of more concrete evidence. Then investigators get wind of research by a former British intelligence officer suggesting Page is acting as a liaison between the campaign and the Russian government. This becomes the critical piece of additional evidence that justifies an initial FISA surveillance application filed in late October.

Horowitz identifies some omissions and questionable characterizations in that initial filing, but none that seem fatal to that first application given the relatively low bar “probable cause” establishes in practice and the limited information available to investigators at the time.

While it’s impossible to know for certain, I believe an initial application free of those defects would still have been approved by the FISC. The Court had already authorized FISA surveillance of Page for reasons unrelated to presidential politics a few years earlier, and former FBI General Counsel. James Baker, who had been working FISA cases since the 1990s, told Horowitz he believed a warrant could have been obtained even without any of Steele’s information. The most glaring problems, in short, come later.

As Horowitz explains, several features of that first application that many observers have understandably found shocking are, by the permissive standards of FISA, not inherently considered “problems” at all—they are how FISA routinely operates.

Most obviously, the application relied critically on a single source, and FBI’s assessment that the source was credible, notwithstanding facts would suggest a potential bias or agenda—Steele had been indirectly commissioned by Democrats to do opposition research on Trump—and even though the FBI had not yet attempted to independently corroborate Steele’s key claims.


But as Horowitz observes, these are not unusual features of a counterintelligence investigation. Intelligence sources—who may include foreign spies or even terrorists—often have agendas, ulterior motives, or axes to grind, and often provide information there is no easy way to directly check. As numerous officials told Horowitz, “the fact that the source information in the FISA application had not yet been corroborated was not unusual in the FISA context.”

Such information is usable as long as the FISC is provided with necessary context and caveats. Thus, at the urging of DOJ attorneys, the application includes an acknowledgement that Steele has been commissioned to conduct opposition research by opponents of the unnamed political candidate Page had been advising.


The “opponents” are not specifically identified as the Democratic National Committee, but this doesn’t appear to have been known to investigators at the time, and the general practice in FISA applications is in any event to obfuscate the names of U.S. persons and groups that are not investigative targets. Trump is referenced only as “Candidate 1” and even the names news outlets in which published articles appeared are omitted.

Similarly, Horowitz explains that nothing in the law or FBI policy requires independent corroboration of information presented in a FISA application. Since any such corroboration would itself have been included in the application, it would have been clear to the FISC that many of the FBI’s crucial assertions hung entirely on their confidence in Steele.

It’s with the renewals that—again, at least by the permissive standards of FISA surveillance—the FBI’s lapses become truly egregious. During the 90 day span of the initial order, the FBI obtains a lot of additional information. They identify and interview Steele’s “primary sub‐​source” and find that his account conflicts with Steele’s reporting in some respects.

The source characterizes as gossip or rumor many of the claims Steele has reported without any such qualification. A formal source review finds that some of Steele’s former intelligence colleagues question his judgment, and that while Steele had provided some useful information for previous investigations, much of it was never independently confirmed.


Review of Page’s older electronic communications finds minimal contact with campaign chair Paul Manafort, though on Steele’s account Page is supposedly acting as an intermediary between Manafort and the Russian government.

By January it should be clear that, at best, many critical claims at the heart of the initial FISA applications require context or qualification; others are simply wrong. None of this makes it into the subsequent renewal applications.

Perhaps the most egregious misconduct occurs in the summer of 2017, when an FBI attorney working on the third and final renewal application alters an e‑mail from a CIA colleague, inserting a false denial that Page had been a CIA source as late as 2013, and provided the agency with information about his contacts with Russian intelligence officers.

The attorney had expressed dismay at the prospect of having to write a “terrible footnote” apprising the FISA Court that such a potentially salient fact was only at this late date being brought to their attention.

By time of the final renewal in June, the FBI’s submissions to the FISC are not merely flawed; they are profoundly misleading, an outrageous dereliction of the Bureau’s duty of candor to the Court. Anyone who claims the Horowitz report as a vindication of the FBI is smoking something.

So what happened? The most conspiratorial version of the “political bias” explanation—the notion that the FBI’s failures represented a conscious plot against Trump—seems fairly easy to dismiss.

Perhaps the most obvious problem with it is simply that there’s just no coherent story on which the protracted surveillance of a peripheral former aide—almost all of it occurring after the presidential election—makes sense as part of a strategy to undermine the new administration.

Nor do the details fit this narrative well: The most serious defects in the renewal applications, for instance, are omissions by case agents that were unknown to the senior officials who reviewed them—a pattern inconsistent with the idea of a vendetta against Trump driven by the Obama‐​appointed leadership.

Instead of dwelling on the countless ways the conspiracy version doesn’t add up, let’s consider the more plausible variant of a “political bias” explanation on which FBI and DOJ officials’ hostility to Trump distorted their judgment, making them too eager to believe the worst of him and his associates, and too willing to overlook facts that contradicted their worst case scenario.

Again, this can’t be ruled out as at least a partial explanation, but there’s very little in the report that suggests it, and a fair amount that cuts against it.

The case for supposing political animus was a major factor rests largely on the now‐​notorious text messages in which Peter Strzok, who played a supervisory role in the early phases of the Page investigation, excoriates Trump and expresses a desire to “stop” him.

But by the time the most egregious omissions and misrepresentations occurred, Strzok was no longer part of the investigation’s chain of command, and even during the earlier phases, was not the one actually drafting FISA applications. Nor does the report depict him as determined to push the limits at any cost: He agreed with other attorneys, for example, that there were insufficient grounds to seek FISA monitoring of another campaign staffer, which the investigative team had proposed.

In any event, as Horowitz explains, “Strzok was not the the primary or sole decision maker on any investigative step in Crossfire Hurricane,” and doesn’t seem to have been personally responsible for any of the individual problems Horowitz documents.


Moreover, the initial Page FISA application, at least, was subject to scrutiny and review by an unusual number of attorneys. Strzok’s bias, in other words, cannot explain the actions of all the other case agents and attorney involved in preparing the application.

Is there evidence that all of them—or at least enough of them—shared his attitude? Horowitz reviewed an enormous number of e‑mails, texts, and instant messages sent by team members and found very little.

The attorney in the general counsel’s office responsible for perhaps the most egregious violation Horowitz found—altering an e‑mail to falsely say Page had not been a CIA source, which affected the final FISA renewal application—expressed extreme dismay at Trump’s election and identified himself with the “resistance.”

Otherwise, however, while Horowitz notes that his office identified a “small number of text messages and instant messages… in which FBI employees involved in the Crossfire Hurricane investigation discussed political issues and candidates,” these “did not raise significant questions of potential bias or improper motivation.”

It’s also worth noting that the counterintelligence investigation of Page was actually opened months prior to the launching of the Crossfire Hurricane probe, by the FBI’s notoriously Trump‐​friendly New York Field Office, for reasons predating his involvement with the Trump campaign.

One reason the investigation focused so intently on Page even before the Steele information was that agents in the New York office had suspicions about his interactions with Russian intelligence officers before he was publicly identified as Trump advisor.

In sum, the initial interest in Page can’t have been motivated by political bias, because it predated Page’s role with the Trump campaign. And if we turn to the later omissions and misrepresentations to the FISA Court, only in one instance do we find evidence of strong political bias against Trump by the person most directly responsible.

That seems insufficient to account for the broader pattern Horowitz found. While, again, it can’t be ruled out as a factor, it’s not an explanation that gets much support from Horowitz’s findings.

Why does that matter? Because if there’s no reason to believe that the defects in the Carter Page FISA process were unique byproducts of anti‐​Trump bias by the officials involved, there’s also no reason to believe failures of this sort are limited to this investigation.

Indeed, to the extent the Horowitz Report does suggest explanations, they point to more general features of FISA investigations.

For instance, consider what officials involved in reviewing some of the later applications told Horowitz about why they failed to identify problems. At least one said that, in general, review of renewal applications was limited to new claims added since the last submission to the Court.

Assertions from previous applications were, in other words, often copied and pasted without being revisited to determine whether they were still supported by the available information.

Another reviewer explains that claims derived form source interviews were “checked” by searching for specific language included in the application to confirm a match—meaning reviewers would not be likely to notice when other portions of the interview contradicted claims in the application, or suggested a need for qualification.

There are also indications in the report that the way agents thought about the justification of continuing FISA surveillance was disconnected from the initial showing to the court in a subtle but important way.

In a criminal inquiry, the purpose of the investigation—obtaining evidence of a crime that can be prosecuted—aligns closely with the showing investigators must make to a court before obtaining a wiretap order: probable cause to believe the will obtain evidence of a crime.

For a FISA order, however, the purpose of obtaining foreign intelligence information is much less tightly connected to the showing that must be made to the court: That the target is (knowingly, in the case of U.S. persons) acting as an agent of a foreign power.

A FISA wiretap might well be regarded as “productive” or “successful” in the sense of yielding information that qualifies as useful foreign intelligence without in any way actually supporting the initial claim that the target is a foreign agent.

After the initial approval, in other words, agents shift from thinking in terms of whether new information supports the original claim that the target is a foreign agent, and instead focus primarily on whether an intercept is producing foreign intelligence.

The simplest explanation for what went wrong here, however, is probably also the most universal: Human beings are subject to confirmation bias. We form theories or beliefs, and then as we gather new information, we tend focus on the data points that support our original theory and ignore or explain away the data points that contradict it.

That’s a primary reason we have warrants evaluated by independent magistrates in the first place: Investigators necessarily form hypotheses and then attempt to prove them, in the process becoming less able to assess whether they’ve done so successfully.

Prosecutors in criminal cases, at least, have a countervailing incentive to think about exculpatory evidence: They know it will count against them if they’ve failed to identify it when their target’s defense attorney is given an opportunity to go over the record. But foreign intelligence surveillance isn’t supposed to yield criminal charges or trials in the vast majority of cases.

Which means intelligence investigations lack the corresponding prospect of getting your knuckles rapped for failing to identify exculpatory evidence. Those are perfect circumstances for confirmation bias to run amok, and that fits what happened here.

Most of the serious defects in the Page application aren’t instances of telling the court something clearly false, but of failing to identify and report information that doesn’t fit the affiant’s theory of the case. 

In one sense, a story where Horowitz’s disturbing catalog of failures, omissions, and misrepresentations can be attributed to “political bias” is comforting. Very few FISA applications involve major political campaigns, so if these problems arose from partisan animosity, we can suppose we would not find a similarly dismaying catalogue in other cases.

If we don’t assume the problems here must be symptoms of political bias—and on the evidence here, it really would have to be nothing more than an assumption—there’s the obvious and disconcerting possibility that you’d find similar patterns in many cases if you began turning over rocks.

Yet as Horowitz himself noted at the Senate Judiciary Committee’s hearing, this is in fact the first time his office has delved so deeply into the progress of a specific FISA order.

In response to a question from Sen. Marsha Blackburn about whether such mistakes were an unusual occurrence, Horowitz had to admit that he simply didn’t know. Perhaps we ought to find out.