The 75-page “Project DEFUSE: Defusing the Threat of Bat-borne Coronaviruses” proposal, submitted by EcoHealth Alliance (PI Peter Daszak) to DARPA’s PREEMPT program on 27 March 2018, requesting $14,209,245 over a planned Dec-2018–May-2022 period of performance. It proposed intensively sampling bat SARSr-CoVs at Yunnan field sites, sequencing their spike proteins, inserting synthesized spike proteins (including “human-specific cleavage sites”) into WIV/SHC014-type bat-CoV backbones, and testing the resulting viruses’ capacity to infect humanized mice, in collaboration with WIV, UNC, and others. DARPA rejected it (see the companion rejection-review node). It is a load-bearing documentary anchor in the lab-leak case because it shows EcoHealth/WIV-linked researchers had, on paper less than two years before SARS-CoV-2 emerged, proposed inserting a human-specific cleavage site into a bat-coronavirus backbone.
relevance_note: Primary documentary evidence that EcoHealth/WIV had, in writing, proposed the specific genetic manipulation (furin-type cleavage-site insertion into a bat-CoV backbone) whose signature is separately debated in SARS-CoV-2’s own genome (see slice D for the sequence argument itself).
Summary (extracted nodes)
A primary document: its documentary facts are extracted as observations resting on the proposal itself (data_basis: [[S-65]]), and its lab-leak implication as one hypothesis plus two arguments (the means/motive/opportunity update, and its intent-not-execution limit). All quotes are transcribed from the scanned PDF’s degraded OCR and reconciled with the widely-circulated verbatim text.
Documentary facts (observations)
O-34 - DEFUSE was a $14.2M EcoHealth-Daszak proposal to DARPA's PREEMPT program submitted 27 March 2018
Provenance/identity of the primary document. Leaked via DRASTIC (Sept 2021), cross-hosted (The Intercept/DocumentCloud/USRTK), never disputed by EcoHealth, and DARPA independently confirmed receiving it. Quotes in this and sibling nodes are transcribed from the scanned proposal PDF, whose OCR is degraded; wording has been reconciled with the widely-circulated verbatim text.
Link to original
O-36 - DEFUSE proposed inserting field-sampled bat spike proteins into WIV1-SHC014 backbones and infecting humanized (hACE2) mice to assess SARS-like disease capacity
The core proposed gain-of-function-style activity: constructing chimeric bat coronaviruses and testing their human-disease potential in hACE2 mice. (Quote reconciled from degraded OCR of the scanned proposal.)
Link to original
O-38 - DEFUSE asserted its chimeric-spike work used bat-SARSr-CoV backbones and was 'exempt from dual use and gain of function concerns'
Documents the consortium’s own advocacy framing that the work fell outside gain-of-function/dual-use oversight — flagged in the source’s motivatedness as risk-minimizing advocacy rather than a neutral biosafety judgment. (Quote reconciled from degraded OCR of the scanned proposal.)
Link to original
O-35 - DEFUSE was a consortium of EcoHealth, UNC (Baric), the Wuhan Institute of Virology (Shi), Duke-NUS, USGS and PARC, with WIV assigned viral testing and humanized-mouse work
Establishes that the WIV was a designated executing partner, and that the chimera/reverse-engineering work leaned on UNC’s Baric group. (Quote reconciled from degraded OCR of the scanned proposal.)
Link to original
O-37 - DEFUSE proposed introducing human-specific (furin) cleavage sites into SARS-related coronavirus spikes
The single most load-bearing documentary fact in the lab-leak institutional case: the consortium proposed, in writing, to insert human-specific/furin-type cleavage sites into SARS-related coronaviruses — the same class of feature separately debated in SARS-CoV-2’s own genome. (Quote transcribed from the scanned proposal’s degraded OCR and reconciled with the widely-circulated verbatim text; the phrase “introduce appropriate human-specific cleavage sites” is the canonical wording.)
Link to original
Hypothesis
H-22 - SARS-CoV-2's furin cleavage site was inserted via DEFUSE-type gain-of-function engineering at the WIV-EcoHealth-UNC consortium
The engineered-origin hypothesis at the granularity DEFUSE supports. The proposal establishes that this precise manipulation was intended at the implicated institutions; whether it was actually executed to produce SARS-CoV-2 is not shown by the document.
Link to original
Arguments
A-46 - DEFUSE shows the exact manipulation debated in SARS-CoV-2's genome was proposed -2 years before emergence at the implicated institutions
reasoning
A striking feature of SARS-CoV-2 is a furin cleavage site absent from its closest known relatives and looking like an insertion. DEFUSE independently establishes that, 20 months before SARS-CoV-2 was first detected in Wuhan, a specific consortium (EcoHealth/Daszak, UNC/Baric, and the WIV/Shi) proposed in writing to do essentially that class of manipulation: screen SARS-related coronavirus spikes for cleavage-site “mismatches” and “introduce appropriate human-specific cleavage sites,” while inserting field-sampled bat spikes into WIV1/SHC014 backbones and testing the chimeras’ human-disease capacity in hACE2 mice. The evidential force is a means-motive-opportunity update: the institutions implicated by the outbreak’s location had the stated intent, the named collaborators, and (via Baric’s reverse-engineering track record and the WIV’s sampling) the capability to build exactly the kind of feature under dispute. This does not observe the virus’s genome (that evidence lives in the genome-structure sources); it raises the base-rate plausibility that such a construct could have originated at these institutions, and thus lifts the prior on the engineered-origin hypothesis well above a naive historical lab-leak rate. Its strength is bounded by the fact that a proposal is evidence of intent, not of a completed construct (see companion argument).
Step 6 verdict — approved (checked)
Reconstructed step: documented intent (the written cleavage-site plan) + capability (Baric’s reverse-engineering record, WIV sampling) + opportunity (outbreak in the WIV’s own city, 20 months later) → the prior on an engineered construct originating at these institutions is higher than a naive historical lab-leak base rate. This is a legitimate means-motive-opportunity Bayesian update, and conditional on the premises (O-35/O-36/O-37) the step goes through. Probed for an undercutting defeater: the natural one is “intent ≠ execution,” but the statement never claims execution — it is hedged to “sharply raises the prior probability,” and that conclusion is immune to the intent/execution gap (which is priced separately by A-47). No defeater breaks the prior-raising link. Traced directly from the premises, so checked.
Link to original
A-47 - DEFUSE is evidence of intent and capability, not of execution- a rejected proposal, chimera work slated for UNC, using known low-risk SARS-1-related backbones
reasoning
Several features of the document cap how far it can support an engineered SARS-CoV-2. (1) It is a forward-looking research plan, not a record of completed work; the proposal was submitted to DARPA and (per the companion rejection node S-66) not funded, so any equivalent work would have had to proceed by another route. (2) The division of labor puts spike reverse-engineering and chimera construction primarily with UNC/Baric’s group (which the proposal itself credits with a “two-decade track record of reverse-engineering spike proteins”), while the WIV is assigned viral testing, binding assays and “some humanized mouse work” — so DEFUSE-as-written does not describe the WIV independently building such chimeras. (3) The backbones named (WIV1, SHC014) are known, characterized, SARS-1-related viruses selected as comparatively low-risk; SARS-CoV-2 is not a close descendant of them, and building it would require deviating substantially from the plan (different, undisclosed backbones; solo execution; a compressed timeline). Consequently DEFUSE functions as a strong prior-raising circumstantial fact — it makes the engineered hypothesis far more than a bare possibility — but it is not itself evidence that the specific construct SARS-CoV-2 was made. The gap between “this was proposed” and “this was done” must be bridged by other evidence.
Step 6 verdict — approved (checked)
Reconstructed step: three features — (1) a forward-looking, DARPA-rejected plan rather than a completed-work record, (2) chimera/reverse-engineering work assigned mainly to UNC/Baric with the WIV assigned testing/assays, (3) named backbones (WIV1, SHC014) being characterized SARS-1-relatives of which SARS-CoV-2 is not a descendant — jointly cap the document’s force to “raises but cannot confirm the specific SARS-CoV-2 construct.” Each sub-inference is valid conditional on the premises (O-35/O-36/O-38, plus the rejection node): a written plan does not entail its execution; a division of labor that does not place solo chimera-building at the WIV means the document does not itself describe the WIV making such a construct; and backbones outside SARS-CoV-2’s lineage mean building it would require undisclosed deviation from the plan. No undercutting defeater survives against the hedged conclusion (it does not deny prior-raising force, only confirmation). Traced from the premises, so checked.
Link to original