Before “Putin preferred Trump” — what the intelligence community's own documents said at the time.
The document: In July 2025 the Director of National Intelligence released a compilation of previously classified primary documents, each stamped “APPROVED FOR RELEASE BY THE DNI ON 17 JULY 2025.” It is 114 pages of internal emails, the pre-election assessment, a pulled Presidential Daily Brief, and the Principals Committee minutes of 9 December 2016.
An honest note: this release is a compilation of primary documents, not a findings report. The word “Obama” does not appear in it. What follows is what the documents themselves say — and every quote is checkable against the file.
Read the documents yourself:
Archived copy of the full PDF
On 31 August 2016, Director of National Intelligence James Clapper wrote that the President had agreed election infrastructure should be treated as critical infrastructure, and that he had “directed my folks to generate an NIE on attendant cyber threats to this key infrastructure, and to get it done sooner than later.” (p. 3)
Within a day, an ODNI official confirmed the change: “Just wanted to make sure, in all this back and forth, that what we'll do, in the interest of time, is an ICA, NOT an NIE.” (p. 2)
Why it matters: An NIE (National Intelligence Estimate) is the community's most formal product, requiring coordination across agencies and an explicit process for recording dissent. An ICA (Intelligence Community Assessment) carries the same weight of authority with far fewer procedural obligations. The record shows the more rigorous standard was set aside to meet a deadline — the procedural shortcut is in the government's own email.
The assessment framed Russia's objective as attacking confidence in the result rather than the result itself: “Foreign adversaries — most notably Russia — probably expect cyber operations to be more effective in undermining the legitimacy of the process than in covertly manipulating the vote outcome.”
And on capability: “We judge that foreign adversaries do not have and will probably not obtain the capabilities to successfully execute widespread and undetected cyber attacks on the diverse set of information technologies and infrastructures used to support the November 2016 US presidential election. We have moderate confidence in our overall threat assessment.” (p. 21)
Why it matters: In September 2016 the community's own position was the opposite of the story that later took hold. The “Putin preferred Trump” judgment did not yet exist.
An August/September 2016 Department of Homeland Security activity report circulated to Clapper's team stated: “The thrust of the analysis is that there is no indication of a Russian threat to directly manipulate the actual vote count through cyber means.” (p. 3, repeated p. 6)
Why it matters: The government knew, in writing, before the election, that the vote count was not the threat. Everything that followed was built on a premise the analysts had already rejected.
On the DNC and DCCC intrusions and the leaking of stolen material, the assessment recorded: “FBI and NSA, however, have low confidence in the attribution of the data leaks to Russia.”
Why it matters: Two of the three agencies would not stand behind the attribution that the public narrative treated as settled fact. FISA surveillance of U.S. citizens followed from sourcing the FBI itself rated low confidence.
During interagency drafting, the Bureau wrote: “we would prefer for the first sentence regarding Russia's intent to be softened. The way it currently reads, it would indicate that we have definitive information that Russia does intend to disrupt our elections and we are uncomfortable making that assessment at this point.” The FBI also asked that a section be retitled “so that it doesn't mislead the reader to believe that the IC currently has information indicating Russia has a known intent to influence the elections.” (p. 8)
Why it matters: The language of the assessment was negotiated in real time — and the participants knew the intent claim outran the evidence.
The drafting record shows the NIC “does NOT want to write on US vulnerability,” and that the paper avoided “the broader influence issue.”
Why it matters: What a product declines to examine is itself a decision. Scope control is how an assessment can be technically accurate and still misleading.
The assessment's “perfect storm” paragraph conceded: “If a 'perfect storm' of coincident political and technological sensitivity were to develop, a cyber adversary might be able to target a small number of critical counties in highly contested states with significant numbers of Electoral College votes. This could potentially alter the apparent outcome of, and almost certainly undermine public confidence in, the election. Although we understand this scenario is unlikely, it remains a possibility that we cannot discount.” (p. 24)
Why it matters: An unlikely scenario, flagged as undiscountable, was the seed from which the vote-changing narrative grew — even though the assessment's own core judgment said the opposite.
The 8 December 2016 PDB, “Cyber Manipulation Of US Election Infrastructure To Remain A Challenge,” assessed: “We assess that Russian and criminal actors did not impact recent US election results by conducting malicious cyber activities against election infrastructure.” On the Illinois voter database, the PDB recorded only “low-to-moderate confidence in the Russian Government's involvement because of our uncertainty about its utility for a state actor.” (p. 44)
Why it matters: After the vote, with the result known, the community's own daily product still said the outcome was not changed. That brief was withdrawn.
The release documents DHS standing up a “centralized bulk screening capability for social media,” to be housed at Customs and Border Protection and the National Counterterrorism Center, with the stated imperatives running from Syrian refugees to Visa Waiver applicants.
Why it matters: This is the machinery High Stakes Treason documents — CBP's TECS entry/exit database, the traveler vetting pipeline, and the screening failures around Tamerlan Tsarnaev and Umar Abdulmutallab — being assembled and formalized. It is the one place where this release and the book describe the same system.
The compilation records an IC ITE pilot, “WAYWARD SKIES,” joining DHS and intelligence-community holdings — travel, immigration and bulk signals data — into a common analytics environment with NCTC and NCSC participating.
Why it matters: The book documents that the NCTC under Brennan held “unfettered access” to information from all 26 of its partner databases, and received simultaneous notification of everything collected. This pilot is that architecture being extended, in the same period.
Read together, the documents show a product whose standard was lowered, whose central claim was hedged, whose attribution was disputed by two of three agencies, whose scope excluded the most relevant questions, and whose later, harder conclusions did not yet exist in September 2016.
The finding that became the public story — that Russia acted to elect Donald Trump — is not in the pre-election record. What is in the record is the process by which such a finding became possible.
All page numbers refer to the PDF as released. Every quotation above is verbatim from that file. The release is a compilation of primary documents; the conclusions drawn here are the author's.
← See also: The Declassified Record — the HPSCI report's findings