At 10:38 on the morning of Feb. 4, 2019, the deputy for United States District Judge Amy Berman Jackson said, “Good morning, Your Honor. This is a sealed proceeding, and the courtroom has been locked.”
It was a hearing at which lawyers for Special Counsel Robert Mueller would argue that Paul Manafort had lied to them during 12 interviews, both before and after striking his plea bargain on foreign lobbying and money laundering charges. We know what was said behind those locked doors because a word-for-word transcript was made of the courtroom proceedings.
Strangely, there was no such transcript of what Manafort had said in his interviews. This made it difficult to reconstruct what his responses to questioning had been, let alone determine whether they were truthful.
But if there was no transcript, surely the court could consult a videotape of his interrogation, the sort of recording that is standard practice for local police who are interviewing, say, a suspected car thief. And if there was no videotape for the judge to watch, surely there was, at the very least, an audiotape.
But no. The bureau has insisted instead, for more than half a century, on summarizing its interrogations, of everyone from witnesses to targets, in a type of memo designated an FD-302 or just “302.” Those memos have been used in recent high-profile investigations to summarize the statements of Manafort, Hillary Clinton, and former national security adviser Michael Flynn, among others. Many were shocked to learn that Clinton’s interview with the FBI about her handling of classified emails was not tape-recorded. They may be more shocked to learn that is standard FBI procedure.
Judge Jackson, who would ultimately side with Mueller’s team on the question of Manafort’s truthfulness, was flummoxed at first. She didn’t quite know what to make of the accusations Manafort had lied after he had pleaded guilty and promised to cooperate. Flummoxed, because she was trying to reconstruct from 302s what had and hadn’t been said.
“That's not the way the 302 read to me when I read it,” she said in response to the prosecution’s characterization of Manafort’s testimony. “This is the problem with not having grand jury testimony, but having to look at a 302.” Clearly frustrated, the judge added, “I may not be able to resolve it on the face of the 302.”
Although some media coverage has suggested otherwise, a 302 is not a transcript. Here’s how a standard FBI interview works. Two agents sit down with the person to be interrogated. One agent does the questioning, the other takes notes — by hand. After the interview is over, back at the bureau office, the agent who took notes uses them to construct a memo summarizing the conversation.
That summary, then, is twice removed from the actual interview. The notes are not an exact rendering of the dialogue; and the memorandum is not an exact copy of the notes. Yet the summary is treated as the official record of the interview. The 302 is a record to which the interviewee is held even though it is not close to a verbatim rendering of what was said — neither questions nor answers.
Those who regularly deal with 302s seem unsurprised when they are missing crucial information. In Jackson’s courtroom, lead Mueller prosecutor Andrew Weissmann was declaring Manafort had been caught in “another false statement.” The proof, said the prosecutor, was found in paragraph 17 of a “declaration” made by an FBI agent involved in the interrogation.
“But it would be reflected in the 302, also?” the judge asked.
“I don’t think it is,” Weissmann responded. “I think it’s only in the declaration.”
The hearing moved on. But let’s pause to consider what just happened. The special counsel’s team was accusing Manafort of a specific lie, but the document ostensibly detailing what Manafort said -- the 302 -- did not have any record of the statement in question. So the prosecutor fell back on a supplemental declaration by an FBI agent. The 302, in other words, was so unreliable that it lacked key information on which prosecutors were basing their case.
At least Manafort’s lawyer, Thomas E. Zehnle, was paying attention. In a post-hearing memorandum, Zehnle and his co-counsels wrote to the court: “Finally, even if one has some concern about how the FBI’s summary report reads, it does not provide the evidentiary basis for finding that Mr. Manafort intentionally misled the investigators.” Pointing to the 302, they added, “The government’s summary is not a grand jury transcript that identifies specific questions and answers; it remains ambiguous.”
By their very nature, 302s are ambiguous not just because they may be unreliable narrations, but because they may include information beyond just what was said. In response to an inquiry from RealClearInvestigations, an FBI official explained, “The FD-302 form, in pertinent part, is used to memorialize interviews as well as other information that may become the subject of testimony.” Asked what exactly “other information” means, or for an example of such information, the official replied, “We’re not able to provide any specifics or further comment for you.”
But Jackson’s courtroom provided some inkling of what “other information” might be. Weissmann called the judge’s attention to “Exhibit 10, page 6, which is the 302” of one of the Manafort interrogations. The prosecutor asked the judge to “look at the 302 — and I'd just like to quote some of it to you, because some of it is factual about what the facts were, not just intuiting what was in someone's head.”
In this inadvertent admission, Weissmann declared that “some” of the government memo being used as evidence “is factual about what the facts were.” And that this part of the memo was “not just intuiting what was in someone’s head.” But doesn’t that suggest that other parts of the 302 passed along information that wasn’t factual or intuited “what was in someone’s head”? A 302 purports to be an accurate account of what was said in an interview. How accurate are agents’ intuitions about their interviewees’ states of mind? Are we willing to convict people and send them to prison for false statements based on what agents intuit?
An Old Habit
These problems are behind a decades-long campaign to have all law enforcement -- from small-town deputy sheriffs to deputy directors of the FBI -- make the recording of interviews standard policy.
One of the leaders of that effort has been civil liberties lawyer Harvey Silverglate, author of the book “Three Felonies a Day: How the Feds Target the Innocent.” He argues that 302s present a pervasive and unnecessary temptation for agents to bend the truth. A “fundamental flaw in the FBI’s truth-gathering apparatus,” Silverglate wrote in 2011, is “the long-defended Bureau-wide policy of not recording interrogations and interviews, a practice that allows the FBI to manipulate witnesses, manufacture convictions, and destroy justice as we once knew it.”
That may sound like strong stuff, unfair even. But the FBI’s own intransigence in the face of the electronic recording movement lends credence to Silverglate’s critique. An FBI response to this proposed reform has become notorious in civil libertarian circles. In 2006, the bureau produced a written rebuttal to the “on-going debate in the criminal justice community whether to make electronic recording of custodial interrogations mandatory.”
The policy memo offered several reasons why the FBI resisted recording, including the telling admission that, when people get a look at FBI interrogations in action, they don’t like what they see: As “all experienced investigators and prosecutors know, perfectly lawful and acceptable interviewing techniques do not always come across in recorded fashion to lay persons as proper means of obtaining information from defendants,” the memo states. “Initial resistance may be interpreted as involuntariness, and misleading a defendant as to the quality of the evidence against him may appear to be unfair deceit.”
The bottom line is that FBI agents feel empowered lie to witnesses or suspects, but when those targets lie to the FBI they are charged with crimes. And being allowed to produce 302s, instead of taped interviews, allows this practice to continue. It’s worth noting that when that memo was produced, the FBI director was Robert Mueller.
The practice of using memos instead of actual transcripts became standard procedure a long time ago, and dates to precedents established by founding FBI Director J. Edgar Hoover. Sometimes the 302 reports reflected noble work. But not always.
In the 1950s, 302 forms were filed by special agents probing the private lives of suspected Communists. In the 1960s, 302 reports were used as a record-keeping tool by FBI agents penetrating the Ku Klux Klan. When Special Agents James P. Hosty Jr. and James W. Bookhout interrogated Lee Harvey Oswald on the day of John F. Kennedy’s assassination, they summarized the questioning in a 302 they dictated the next day. It was a measure of the artificiality of the memos, and the agents’ skewed sense of what was important, that they stated in the first paragraph of their report that Oswald “made many uncomplimentary remarks about the FBI.”
During the Watergate scandal, 302s played a crucial role—a role that allowed for manipulation and misdirection. Deputy FBI Director Mark Felt filed a memo suggesting leaks to The Washington Post were coming from a Miami prosecutor who had access to 302 reports. Yet it was Felt who was Bob Woodward’s infamous “Deep Throat.”
There are more recent examples. Immediately after assuring President Trump in private meeting, “I don’t do sneaky things, I don’t leak, I don’t do weasel moves,” then-FBI director James Comey went to his car, got out his laptop and immediately began writing a 302-style memo. Which he later leaked. According to his own congressional testimony, Comey did the same thing to former White House Chief of Staff Reince Priebus. Yet Comey seemed surprised anyone would see those as weasel moves or sneaky things, the obvious irony being that we only know about the “weasel moves” quote because Comey put them in his 302-type memo.
“The Bureau cherishes its procedures and lives by them,” Andrew McCabe, another fired top FBI official, wrote in his recent memoir. “The official write-up on my own application interview, conducted in Philadelphia more than two decades ago, can be found somewhere in the personnel files at the J. Edgard Hoover Building. It’s a form known as a FD-302. Every interview conducted by an FBI agent is reported or summarized on a 302. This form is the most basic building block of an investigation.”
Not all FBI agents believe they should be. Around the same time as the controversial “misleading a defendant” explanation was being promulgated, Special Agent Brian Boetig co-authored an article, “Revealing Incommunicado: Electronic Recording of Police Interrogations,” for the FBI Law Enforcement Bulletin. In it, he and his fellow authors made the case for recording all interviews. (Boetig is not squeamish: His crime-fighting bona fides are solid enough that he was recently promoted to be assistant director of the bureau’s Weapons of Mass Destruction Directorate.) “Testimony regarding what transpired inside the interrogation room,” they wrote, “can become tainted if only the participants witnessed what occurred.”
Boetig and his co-authors argued that when interviews and interrogations aren’t recorded any number of things can go wrong: “First, problems associated with recollection can contribute to conflicting statements. Interrogations often last for hours and exact transcripts cannot precisely memorialize everything. Furthermore, a trial may not occur for years after the interrogation, reducing the ability to cognitively recall all of the specific details and circumstances not recorded in notes or reports.”
Beyond that, note jotters and memo writers can simply get things wrong. It would be a remarkable feat of note-taking to capture every word that is said in a fluid conversation; so to start with, pertinent dialogue is missing even in the most scrupulously produced 302. There may also be things that the note taker mishears, misunderstands, or scribbles down incoherently.
Again, remember that the notes aren’t the finished product: They are used as a reference by the agent as he or she — perhaps days later — reconstructs the interview in summary form. Such summaries might be useful for agents and case managers to keep track of investigations, but the use of 302s goes far beyond that. They become the official record of what was said in an interview. If one contradicts that record in subsequent interviews or testimony, the 302 can be used as evidence that one has lied to the government.
The Michael Flynn Case
It’s one thing to be held accountable for what one says in a tape-recorded interview, where there is a complete record of what was said on both sides of the conversation. But how confident can we be that a memo summarizing handwritten notes is accurate, even assuming agents’ best good-faith efforts? Confident enough to send someone to the penitentiary for many years?
Consider Michael Flynn, who pleaded guilty to “making materially false statements and omissions” to FBI agents because the 302s of his interview did not match other evidence the bureau had collected. It turns out that mistakes were made -- and not only by Flynn.
For starters, one of the two FBI agents who interviewed Flynn on Jan. 24, 2017 was Peter Strzok, who was later removed from the investigation because of deep anti-Trump animus he expressed in texts with a fellow FBI employee. Even then, according to congressional testimony, Strzok and the other agent didn’t really believe Flynn lied to them. And, of course, they had no tape recording to help them weigh the facts. As McCabe himself wrote in a memo, “[I]t was not a great beginning of a false statements case.”
That aspect of the saga received some media attention. But another one did not. Late last year, District Court Judge Emmet G. Sullivan ordered the special counsel to release the FD-302 memorializing the interview with Flynn. On Dec. 17, the Mueller team complied: “the government hereby files two redacted versions of the FD-302 report summarizing the FBI’s interview of the defendant on January 24, 2017.”
The first had been originally filed Feb. 15, 2017, three weeks after the interview with Flynn. Three weeks seems an awfully long time for recollection, especially of subtle details that may lead to a man being prosecuted for lying to the FBI. Is three weeks common or even consistent with official policy? It’s hard to say, because the publicly available version of the FBI’s Domestic Investigations and Operations Guide reads, “Any matter that may be testimonial must be documented using an FD-302 within [REDACTED].”
Let’s put aside the question of how it could possibly be the case that the standard deadline for filing a 302 is some sort of secret that warrants redaction. Let’s assume that three weeks is within the window. But even if three weeks doesn’t seem like a long time, that’s nothing compared to how long it took to file the second, final version of the Flynn 302 — 3½ months. Here’s how Robert Mueller explained what happened: “The content of both versions of the report is identical, except that the first version, which was digitally signed and certified in February 2017, inadvertently contained a header labeled ‘DRAFT DOCUMENT/DELIBERATIVE MATERIAL.’ Once that error was recognized,” Mueller said, “the header was removed and a corrected version, omitting only the header, was re-signed and re-certified in May 2017.”
How did the words “DRAFT DOCUMENT/DELIBERATIVE MATERIAL” find their way onto the 302? And don’t those words suggest that the document was still open to revision? The one thing we can be quite sure of, however, is that inaccurate claims were made in an official FBI record of an interview: Isn’t that the sort of thing that gets people in trouble?
But let’s be more generous than aggressive prosecutors tend to be and accept that the error was inadvertent, as Mueller claims. That would be proof of an astonishing tendency to error in the creation of FD-302s. Here we have a crucial 302 of the highest importance -- one providing the pretext to prosecute the White House national security adviser. And it not only has an “error” at the top of the page, that “inadvertent” text is centered and written in bold, all-capital letters. If that’s the sort of supersized mistake that happens in a high-priority case, what does that tell us about the accuracy and reliability of the average FD-302?
The concern isn’t a new one: “You can have a conversation with an agent,” Robert Kennedy’s press secretary when RFK was attorney general, once told journalist Victor Navasky, “and when it is over he will send a memo to the files. Any relation between the memo and what was said in the conversation may be purely coincidental. You would think you were at different meetings.”
It was that sort of reputation that led the Department of Justice in 2014 to issue a new Policy Concerning Electronic Recording of Statements. Promulgated by then-Deputy Attorney General James M. Cole, the document opens with this declaration: “This policy establishes a presumption that the Federal Bureau of Investigation (FBI), the Drug Enforcement Administration (DEA), the Bureau of Alcohol, Tobacco, Firearms, and Explosives (ATF), and the United States Marshals Service (USMS) will electronically record statements made by individuals in their custody.”
There was great anticipation that the new policy would finally be the end of commemorating interviews in memos derived from handwritten notes. “This policy change is an important step in the right direction,” the Harvard Law Review declared in 2015, “reflecting a growing movement that has recognized the benefits of recording interviews; however, the new policy puts in place little express accountability for failure to comply with the presumption.”
Not only is there “little express accountability” for not recording interviews at the FBI, the bureau has made it policy to interpret the presumption of recording in the narrowest possible way, while obliging agents to file 302s as a matter of course. DOJ policy notwithstanding, “FBI interviews with witnesses are rarely recorded,” says one prominent white-collar Washington, D.C., defense lawyer. “About 99 percent of the time agents take notes during the interview and then turn those notes into a 302.”
That 99 percent may be a slight exaggeration, but what is the real ratio of taped to typed interviews? If the bureau knows, it isn’t willing to share: “The FBI does not maintain statistical information about the aggregate number of electronic recordings or FD-302s prepared in a given window of time,” an FBI official told RealClearInvestigations.
Asked why 302s are still standard operating procedure, the bureau is coy. “The electronic recording of non-custodial interviews is not required,” the same FBI official said in an email exchange. But that covers a lot of suspects and witnesses, as any number of FBI targets can attest, including Michael Flynn. Although the current DOJ written guidelines call for video or audio recording of interviews of persons under arrest, it also includes this caveat: “Interviews in non-custodial settings are excluded from the presumption.” And “non-custodial” is an interesting term of art.
“There is a presumption that statements made by persons in FBI custody must be recorded,” the bureau instructs agents in the 2016 edition of the Domestic Investigations and Operations Guide. But that leaves a lot of running room to interrogate people without cameras or recording devices. To the FBI, “custody” only applies “following arrest and prior to initial appearance when the arrestee is in a place of detention with suitable recording equipment.”
That means for those who have not yet been arrested when they are interviewed-- as was the case with Lt. Gen. Flynn -- FBI policy is to rely on 302s to memorialize what was said and done. The same holds true of those who have been arrested and then have been arraigned, which explains why the special counsel’s office and the FBI didn’t feel obliged to record any of their dozen interviews with Manafort.
The FBI does allow agents to consider recording non-custodial interviews, but requires “the interviewing employee” to consider a range of factors. One such factor is “[w]hether the interviewee’s own words and appearance (in video recordings) would help rebut any doubt about the meaning, context or voluntariness of his/her statement or confession raised by his/her age, mental state, educational level, or understanding of the English language; or is otherwise expected to be an issue at trial, such as to rebut an insanity defense; or may be of value to behavioral analysts.”
The other factors similarly focus on the question of whether a recording will help get a conviction rather than whether it will help get at the truth.
Recordings are not always perfect representations of what happened in interviews, either. (There is, for example, literature on how camera placement can affect viewers’ interpretations of a subject’s demeanor.) But electronic recording is clearly superior to the 302 process with its invitation to error and risk of baked-in bias. Yet, in a society where any and all of one’s movements are captured on videotape (just ask Jussie Smollett), the one situation in which there is likely not to be a camera running is in an official FBI interview.
What will it take to have the FBI give up its stubborn commitment to the scribble-and-type method of memorializing interviews? A decree from the attorney general or the president would likely get the job done. Or legislation from Capitol Hill, which controls the FBI’s funding. Concern about retaining the public’s confidence in the bureau certainly hasn’t been enough. Nor have humiliating courtroom defeats.
A year ago, a jury in Florida acquitted Noor Salman of all charges against her in the massacre at the Pulse nightclub in Orlando. The wife of Omar Mateen, the mass murderer who carried out the attack, Salman was charged by federal authorities with obstruction of justice and aiding and abetting her husband in the attack.
The Justice Department’s case was based almost solely on the confession FBI agents got Salman to sign after questioning her for 18 hours without a lawyer – and, crucially, as it turned out -- without a tape recorder. At the end of that interrogation, she signed a statement based on information fed to her by the two investigating FBI agents, a statement the agents wrote themselves. Some of that information would be disproven at trial, meaning that the prosecution relied on a “confession” that contained information fed to her by FBI agents that the government subsequently found out to be untrue. Moreover, at the trial, the agents contradicted each other on significant portions of what she said. These discrepancies might have been resolved by a tape or transcript of the interview, but of course, nothing of the kind existed.
“The FBI must join the rest of law enforcement and record all statements,” one of Salman’s attorneys, Charles Swift, said after the acquittal was announced. “It’s ridiculous if they don’t.”
The jury foreman agreed. In a statement issued afterward to the media, the foreman said the jury was convinced that Noor Salman did know what her husband was planning to do. But based on the incomplete method the FBI used to collect evidence, they couldn’t in good conscience convict her.
“I wish that the FBI had recorded their interviews with Ms. Salman,” the foreman added, “as there were several significant inconsistencies with the written summaries of her statements.