# Gospel (HaBsora) AI Targeting System ## BLUF **Gospel** (Hebrew: *HaBsora*) is an AI-assisted targeting system operated by the Israel Defense Forces that generates **infrastructure and building targets** for airstrikes during operations in Gaza. Where [[01 Actors & Entities/14_Corporations_&_Tech/Lavender|Lavender]] designates individual humans as targets, Gospel designates physical structures and sites — homes identified as belonging to Hamas operatives, buildings used as operational infrastructure, and facilities assessed as militarily significant. Gospel is the less publicized but arguably more operationally important of the IDF's AI targeting systems: it has produced target recommendations at rates that exceed human strike capacity, inverting the traditional bottleneck in air operations from *finding targets* to *servicing the target list*. --- ## Operational Function Gospel is designed to accelerate the **target generation** phase of the kill chain by algorithmically identifying and recommending physical targets based on multi-source intelligence fusion: **Input sources (assessed):** - Signal intelligence (SIGINT) — communications intercepts associating individuals with specific locations - Geospatial intelligence (GEOINT) — commercial and military satellite imagery, drone surveillance - Human intelligence (HUMINT) — informant and captive reporting - Open-source intelligence (OSINT) — social media analysis, leaked documents - Pattern-of-life analysis — behavioral signatures associated with operational facilities **Output:** - Ranked target recommendations with confidence scores - Associated rationale (why this building is assessed as a legitimate military target) - Estimated collateral damage projections (civilians assessed to be present at strike time) - Weapon-to-target recommendations (munition type based on target hardness) **Throughput:** Investigative reporting (+972 Magazine, November 2023) described Gospel as having transformed the IDF's targeting capacity from approximately 50 targets per year for a traditional staff cell to "hundreds of targets per day" during active Gaza operations. --- ## The "Mass Assassination Factory" Reporting The November 2023 +972 Magazine investigation, "A mass assassination factory: Inside Israel's calculated bombing of Gaza," provided the first detailed public account of Gospel. Key findings: ### Target Categories Sources described Gospel generating four primary target categories: 1. **Tactical targets** — structures with direct military function (observation posts, weapons storage, tunnel entrances) 2. **Underground targets** — suspected tunnel networks, often targeting entire buildings atop assessed tunnel routes 3. **Power targets** — buildings chosen for symbolic/psychological effect, assessed as associated with Hamas but also calibrated to produce collapse-of-will effects (high-rise buildings, government offices, banks, universities) 4. **Family homes** — residences of individuals assessed as Hamas operatives, typically struck when the operative was at home (overlapping with [[01 Actors & Entities/14_Corporations_&_Tech/Lavender|Lavender]]/Where's Daddy doctrine) ### Civilian Damage Calibration Sources described institutional authorization of specific civilian casualty thresholds: - Low-ranking Hamas member's home: up to 15–20 civilian deaths acceptable per strike - Senior Hamas commander: no upper limit specified - Power targets: damage acceptable if the target's assessed propaganda/collapse effect justified it These thresholds were not case-by-case proportionality analyses under IHL; they were categorical authorizations applied to target classes. ### Bottleneck Inversion Traditional military air operations have a target generation bottleneck: intelligence staff cannot produce high-quality target packages fast enough to keep strike aircraft fully employed. Gospel inverted this: the system produced target packages faster than the IDF could strike them. The bottleneck became strike capacity (available aircraft, munitions stocks) and human-in-the-loop approval — not target identification. --- ## Relationship to Lavender and Where's Daddy The three systems operate as an integrated targeting architecture: | System | Targets | Primary Function | |---|---|---| | [[01 Actors & Entities/14_Corporations_&_Tech/Lavender\|Lavender]] | Individual humans | Designates persons as Hamas operatives | | Where's Daddy | Individual humans (tracking) | Locates designated targets at residential locations for strike timing | | **Gospel** | Physical structures | Designates buildings and infrastructure as military targets | The systems share intelligence infrastructure and feed each other: a person designated by Lavender may have their home added to Gospel's target list; a building designated by Gospel as operationally significant may drive Lavender to designate its occupants. This integration produces the **algorithmic kill chain** in its mature form: targets (persons and places) are generated, tracked, and approved at machine speed, with human review compressed to seconds per decision under operational tempo. --- ## Analytical Assessment ### What Gospel Represents Technically Gospel is a **decision support system**, not an autonomous weapon. It recommends targets; humans approve strikes. This distinction matters legally but increasingly less operationally — at throughput rates of hundreds of target approvals per day by junior officers, the meaningful human control element is structurally compressed. Technically, Gospel appears to be an application of general machine learning techniques — pattern recognition, natural language processing of intelligence reports, predictive modeling — to the specific problem of military target identification. It is not a qualitatively novel AI capability; it is the industrial application of established ML techniques to a high-stakes military domain. ### What Gospel Represents Doctrinally Gospel is the operational validation of the **kill chain compression** doctrine originally theorized in US [[03 Weapons & Systems/Cyber Capabilities & Tools/Project Maven and Kill Chain Compression|Project Maven]] programs. The theoretical claim — AI compresses the analysis phase; humans serve as approvers at machine tempo — has been tested at operational scale in Gaza. **Doctrinal observation:** At operational tempo, the meaningful constraint on operations becomes **strike capacity** (aircraft availability, munition stocks, political authorization for additional strikes), not intelligence quality or targeting rigor. This is a qualitatively different operational environment than pre-AI targeting. ### What Gospel Reveals About IHL Compatibility The fundamental IHL requirements — distinction, proportionality, precaution — assume that targeting decisions can be made with deliberate consideration of specific circumstances. Gospel at operational tempo makes this structurally difficult: - **Distinction:** Algorithmic designation based on pattern-of-life signatures does not constitute the individualized combatant-vs-civilian assessment IHL requires - **Proportionality:** Categorical civilian casualty thresholds applied at target-class level are not proportionality analysis - **Precaution:** Striking family homes to maximize probability of target presence does not minimize civilian harm These are not Gospel-specific questions; they are structural questions about AI-assisted targeting at operational scale, which Gospel is the most extensive active case study of. --- ## International Response and Replication Gospel's operational record is being studied by every military currently developing AI targeting capability: - **United States:** The Maven Smart System (MSS), also operated by Palantir with IDF, shares intellectual and technical heritage with Gospel - **NATO allies:** UK, France, Australia actively monitoring; doctrine developing in NATO's Joint Intelligence, Surveillance and Reconnaissance architecture - **China:** PLA AI research appears to be advancing comparable systems, though operational deployment status is unclear - **Russia:** Limited public evidence of deployed systems; high likelihood of parallel development under wartime pressure The precedent Gospel has established — that AI-assisted targeting at operational scale produces orders-of-magnitude higher target throughput, with civilian casualty rates that are politically sustainable under specific conditions — will be studied and operationally replicated regardless of international legal concerns. --- ## Key Connections - [[01 Actors & Entities/14_Corporations_&_Tech/Lavender]] — the individual-human targeting counterpart - [[07 Current Investigations/Active Investigations/The IDF's Kill Machine]] — the investigation dossier covering all three systems - [[03 Weapons & Systems/Cyber Capabilities & Tools/Project Maven and Kill Chain Compression]] — US doctrinal predecessor - [[01 Actors & Entities/14_Corporations_&_Tech/Palantir Technologies]] — AI infrastructure layer - [[02 Concepts & Tactics/Kill Chain]] — the framework being compressed - [[02 Concepts & Tactics/Algorithmic Warfare]] — the doctrinal category - [[02 Concepts & Tactics/Algorithmic Targeting Systems (2022-2024)]] — the broader phenomenon - [[04 Current Crises/Active Conflicts/Gaza War]] — the operational context - [[06 Authors & Thinkers/Contemporary Analysts/Jenna Jordan]] — strategic critique of high-throughput decapitation doctrine