BDS-1000 Dossier: OpenAI, Inc.

Target Profile

Executive Summary

OpenAI’s documented nexus to Israel/Palestine operates almost entirely through its strategic partnership with Microsoft. The company’s GPT-4 language models are accessed by Israeli military, intelligence, and civilian-administrative entities exclusively via Microsoft’s Azure OpenAI Service, which serves as the sole delivery mechanism for OpenAI’s technology to Israeli government end-users. Journalistic investigations confirmed that at peak periods in 2024, GPT-4 accounted for approximately one-quarter of the Israeli military’s total machine learning tool consumption on Azure, with consumption increasing approximately 64-fold by March 2024 compared to pre-war baselines ¹².

Multiple journalistic investigations documented OpenAI’s GPT-4 models integrated into the Israeli military’s AI-assisted targeting infrastructure, including the Lavender system (AI targeting for air strikes), The Gospel system (mass target generation), and the “Where’s Daddy?” tracking system ¹²³. Azure also hosts population control systems including “Rolling Stone” (Palestinian population registry) and “Al Munaseq” (movement permit management) ⁴. Microsoft confirmed in September 2025 that it suspended specific Azure services to Israeli military Unit 8200 following evidence of mass surveillance of Gaza and West Bank civilians ⁵.

OpenAI has no direct contracts, subsidiaries, or physical presence in Israel. The company has consistently stated it does “not have a partnership with the IDF” — a claim verified as accurate regarding direct contractual relationships ¹. However, all confirmed access flows through Microsoft’s Azure platform under Microsoft’s separate $133 million IMOD contract. The company’s January 2024 policy change removing explicit prohibition of “military and warfare” uses coincided with the period of dramatically increased Israeli military consumption of GPT-4 via Azure ¹³.

The BRS score of 497 places OpenAI in Tier C (High), driven primarily by the V-MIL score of 6.43 reflecting confirmed military application of the company’s technology through a third-party platform, balanced against the absence of direct contractual relationships or physical operational presence in the region.

Timeline of Relevant Events

Corporate Overview

OpenAI was founded in 2015 as a Delaware nonprofit corporation with the stated mission of ensuring artificial general intelligence benefits humanity. The corporate structure evolved through OpenAI LP (2019), and the company completed a restructuring to a Public Benefit Corporation in October 2025. The current ownership structure following the PBC conversion shows OpenAI Foundation holding 26%, Microsoft holding 27%, and employees/other investors holding 47% ⁸.

Subsidiaries and Related Entities:

• OpenAI Global, LLC — Primary operating entity for international operations
• OpenAI Public Sector LLC — Registered December 2024, holds U.S. federal government contracts
• Safe Superintelligence Inc. (SSI) — Separate company co-founded by former OpenAI Chief Scientist Ilya Sutskever in June 2024, with research facilities in Tel Aviv and Palo Alto. SSI is not a subsidiary or affiliate of OpenAI ¹⁰¹¹.

Microsoft Relationship:

Microsoft is OpenAI’s exclusive cloud infrastructure partner for commercial API deployment, having invested approximately $13 billion in OpenAI. All GPT-4 access for Israeli military customers flows through Microsoft’s Azure OpenAI Service under Microsoft’s separate $133 million contract with the Israeli Ministry of Defence ¹².

Israeli Entities and Franchise Relationships:

No public evidence identifies OpenAI maintaining subsidiaries, franchise operations, or authorized reseller relationships within Israel or occupied territories. OpenAI’s technology reaches Israeli end-users exclusively through Microsoft’s Azure platform.

Domain Summaries

V-MIL: Military

Mechanism of Involvement

OpenAI’s mechanism of involvement with the Israeli military is indirect and mediated entirely through Microsoft’s Azure cloud platform. OpenAI Public Sector LLC holds a $200 million ceiling contract with the United States Department of Defense for “frontier AI capabilities” in warfighting and enterprise domains through July 2026 ¹³¹⁴. However, no direct contract between OpenAI and the Israeli Ministry of Defence, IDF, Israel Prison Service, or Israel Border Police has been identified in public evidence ¹³.

The confirmed pathway is: OpenAI provides GPT-4 to Microsoft Azure → Microsoft Azure OpenAI Service → Israeli military end-users under Microsoft’s $133 million IMOD contract. Internal Microsoft data reviewed by AP News and The Guardian showed Azure AI usage by the Israeli military increased approximately sevenfold in the first month of the Gaza war and approximately 64-fold by March 2024 compared to the pre-war baseline in September 2023 ¹²³. At one stage in 2024, OpenAI’s tools accounted for approximately one-quarter of the Israeli military’s machine learning tool consumption on the Azure platform ¹³.

Multiple journalistic investigations confirmed that GPT-4 was consumed by the Israeli military’s AI-assisted targeting pipeline, including the Lavender system (AI targeting for air strikes), The Gospel system (AI system generating bombing targets at scale), and the “Where’s Daddy?” system used for tracking and striking individuals ¹²³. Azure also hosts the “Rolling Stone” system (population and movement registry for West Bank and Gaza) and the “Al Munaseq” app (Palestinian movement permit management) ⁴.

Col. Racheli Dembinsky, commander of the IDF’s MAMRAM unit, publicly stated that cloud providers gave the IDF “very significant operational effectiveness” during the Gaza ground invasion, with Azure logos appearing alongside AWS and Google Cloud in her official presentation slides ¹.

Counter-Arguments and Evidence Limits

OpenAI’s strongest defense rests on the absence of a direct contractual relationship with Israeli military or security entities. The company has stated publicly that it does “not have a partnership with the IDF” — a claim verified as accurate with respect to direct contractual relationships ¹. All confirmed access to OpenAI models by Israeli military end-users is mediated through Microsoft’s Azure platform rather than through direct OpenAI contracting.

OpenAI can argue that:

1. No direct contracts: No public evidence identifies a direct contract, tender award, or memorandum of understanding between OpenAI and Israeli Ministry of Defence, IDF, Israel Prison Service, or Israel Border Police ¹²³.

2. Platform intermediary: Microsoft Azure operates as an independent intermediary; OpenAI does not control which end-users Microsoft serves or monitor downstream consumption of its models on Azure.

3. Policy framework: OpenAI published National Security Partnership Guidelines in October 2024 establishing three “red lines”: no use for mass domestic surveillance, autonomous weapons systems, or high-stakes automated decisions ¹⁴. The company disrupted an Israeli government-funded influence operation (STOIC) in May 2024 ⁷⁸.

4. No settlement presence: OpenAI does not appear in the UN OHCHR settlement enterprise database (September 2025 iteration, 158 companies) ¹⁷.

5. Service suspension: Microsoft, not OpenAI, suspended specific services to Israeli military Unit 8200 in September 2025 following a review of mass surveillance usage ⁵.

However, the audit notes that OpenAI’s January 2024 policy change removing explicit prohibition of “military and warfare” uses preceded the dramatic increase in Israeli military consumption of GPT-4 via Azure, and the company has not issued public statements specifically addressing civilian harm allegations regarding its technology’s role in Israeli military operations ¹³¹⁸.

Named Entities and Evidence Map

V-DIG: Digital

Mechanism of Involvement

OpenAI’s digital infrastructure nexus to Israel operates entirely through its dependency on Microsoft Azure. OpenAI has no independent cloud infrastructure; all commercial API deployment runs through Microsoft’s infrastructure ¹¹⁵. This creates a structural channel through which Israeli military and government customers access OpenAI models.

Microsoft operates Azure cloud regions in Israel, launching its first Israeli Azure cloud region in November 2023 ¹⁹. Microsoft is classified as an S500 (top priority) client for the Israeli military and holds a $133 million, three-year contract with the Israeli Ministry of Defense dating from 2021 ¹². Microsoft engineers collaborated directly with Israeli military intelligence, including on-site at army bases, to customize Azure implementation for Unit 8200 and Unit 9900 requirements ¹²⁰.

Israeli military AI targeting systems (Lavender, The Gospel/Habsora, Where’s Daddy?) are documented in UN and NGO reporting, with GPT-4 used as a supporting tool for translation, transcription, and document analysis ⁹¹⁴. However, no public evidence confirms that GPT-4 is directly integrated into the targeting decision pipeline rather than used for supporting functions ⁹¹⁴.

Counter-Arguments and Evidence Limits

OpenAI’s strongest digital-domain defenses include:

1. No direct Israeli infrastructure: No public evidence identifies OpenAI operating, leasing, or co-locating data center infrastructure within Israel ¹¹⁵.

2. No direct vendor relationships: No public evidence identifies OpenAI holding licensing, subscription, or integration relationships with Israeli-origin technology vendors including Check Point, Wiz, SentinelOne, CyberArk, or Palo Alto Networks ¹⁵.

3. No surveillance technology integration: No public evidence identifies OpenAI directly using facial recognition, biometric identification, or behavioral analytics technologies of Israeli origin ¹⁵.

4. Disruption of abuse: OpenAI proactively disrupted the STOIC influence operation run by Tel Aviv-based actors using its models to generate pro-Israel content ⁷⁸.

However, the audit notes that OpenAI’s total dependency on Microsoft Azure means the company has no independent ability to restrict access to its models by Israeli military end-users once Microsoft makes them available on the Azure platform. The company’s policy change in January 2024 removing “military and warfare” prohibitions preceded the period of dramatically increased Israeli military consumption ⁶.

Named Entities and Evidence Map

V-ECON: Economic

Mechanism of Involvement

OpenAI’s economic nexus to Israel operates in the reverse direction from many other companies: Israeli military and government entities are consumers of OpenAI’s services via Microsoft Azure, rather than OpenAI sourcing inputs from Israeli suppliers. No public evidence identifies OpenAI directly procuring physical goods from Israeli suppliers or maintaining operational presence in Israel.

The economic linkage operates through three channels:

1. Microsoft ownership stake: Microsoft holds approximately 27% of OpenAI post-restructuring, representing an estimated $135 billion valuation on a committed investment of approximately $13 billion ⁸. This creates an economic interest linking OpenAI’s financial performance to Microsoft’s broader commercial activities, including Microsoft’s Israeli subsidiary operations.

2. Sam Altman personal investment: Sam Altman invested in Apex Security, an Israeli cybersecurity startup, via his personal investment vehicle Alt Capital, participating in a $7 million seed round in 2024 ¹⁷¹²²¹. Apex Security was co-founded by veterans of IDF Unit 8200. Altman exited this investment when Tenable acquired Apex in May 2025 at a valuation exceeding $105 million ¹⁷²¹.

3. Technology consumption: Israeli military and government entities consume OpenAI’s models through Microsoft’s Azure platform, generating revenue that flows to OpenAI (via Microsoft’s licensing arrangement) and ultimately to shareholders including Microsoft.

No public evidence identifies OpenAI maintaining wholly-owned subsidiaries, joint ventures, or dedicated import entities within Israel or occupied territories. OpenAI does not appear in the UN OHCHR settlement enterprise database ¹⁷.

Counter-Arguments and Evidence Limits

OpenAI’s strongest economic defenses include:

1. No direct investment in Israel: No public evidence identifies OpenAI making direct capital investments — acquisitions, data centers, real estate holdings, or logistics hubs — within Israel or occupied territories ²²²³.

2. No physical presence: No public evidence identifies OpenAI operating any office, sales operation, support center, warehouse, or retail location within Israel or occupied territories.

3. No Israeli revenue disclosure: No public evidence identifies OpenAI disclosing revenue generated from or attributed to Israel as a market.

4. Reverse economic flow: The economic integration runs in the opposite direction from typical settlement complicity — Israeli entities are consumers of OpenAI’s services, not suppliers.

However, the audit notes that Microsoft’s 27% ownership stake creates an indirect economic linkage to Microsoft’s Israeli operations, and Sam Altman’s personal investment in an Israeli defense-linked startup represents individual (not corporate) exposure to Israeli economic activity.

Named Entities and Evidence Map

V-POL: Political

Mechanism of Involvement

OpenAI’s political nexus to Israel/Palestine manifests through corporate communications, executive activities, governance composition, and policy positions:

1. Executive engagement: Sam Altman visited Israel in June 2023, meeting President Herzog and Prime Minister Netanyahu, stating Israel would have a “huge role” in AI development ⁶. No comparable public statements by Altman specifically addressing the Gaza conflict or civilian harm have been documented.

2. Board composition: Paul M. Nakasone, retired U.S. Army General and former NSA Director (2018–2024), was appointed to OpenAI’s Board of Directors in June 2024, joining the Safety and Security Committee ⁹.

3. Policy change: OpenAI revised its Usage Policy in January 2024, removing the explicit prohibition on “military and warfare” uses, replacing it with narrower language prohibiting only “weapons” development ⁶.

4. Influence operation disruption: OpenAI disrupted an Israeli government-funded influence operation (STOIC) in May 2024 that used OpenAI models to generate pro-Israel content ⁷⁸.

5. Lobbying: OpenAI’s federal lobbying spending increased from $260,000 in 2023 to $1.76 million in 2024, representing a nearly sevenfold increase ¹². No evidence identifies OpenAI lobbying specifically on anti-BDS legislation or Israel trade agreements.

6. No BDS targeting: No organized boycott, divestment, or sanctions campaign specifically targeting OpenAI for Israeli technology relationships was identified ¹⁸²⁵.

Counter-Arguments and Evidence Limits

OpenAI’s strongest political defenses include:

1. No corporate statements on conflict: OpenAI has not issued official corporate statements specifically addressing the Israel-Palestine conflict, maintaining a position of neutrality on the matter.

2. Disruption of pro-Israel influence operation: OpenAI proactively identified and disrupted the STOIC operation, demonstrating willingness to enforce usage policies against Israeli state-affiliated actors ⁷⁸.

3. No BDS campaigns: OpenAI has not been the subject of documented organized BDS campaigns, unlike Microsoft and Google who face sustained activist pressure.

4. No physical presence: No evidence identifies OpenAI maintaining offices or operations in occupied territories.

5. Defense policy guardrails: OpenAI published National Security Partnership Guidelines establishing “red lines” against mass surveillance and autonomous weapons ¹⁴.

However, the audit notes that the January 2024 policy change removing “military and warfare” prohibitions — coinciding with the period of increased Israeli military consumption — and the appointment of a former NSA Director to the board, create perception of alignment with defense-sector interests regardless of corporate intent.

Named Entities and Evidence Map

BDS-1000 Score (V4)

• V_MAX: 6.43
• Sum_OTHERS: 7.58
• BRS Score: 497
• Tier: C (High)

The V-MIL score of 6.43 drives the BRS calculation, reflecting confirmed integration of OpenAI’s GPT-4 models into Israeli military targeting and surveillance infrastructure through Microsoft’s Azure platform. The V-DIG score of 4.50 reflects the structural dependency on Microsoft Azure as the exclusive delivery mechanism. The V-ECON score is minimal (0.05) given the absence of direct investment or operational presence. The V-POL score of 3.03 reflects executive engagement and governance composition without direct political advocacy or BDS targeting.

The methodology uses scale-free Impact (activity type), Magnitude (scale), and Proximity (directness) to compute domain scores, with final BRS derived from the highest domain score plus half the sum of others. All scores are evidence-only, derived from the four domain audits, and reflect human-vetted findings.

Methodology Note

• Evidence-only approach: All claims trace to findings from the four domain audits (V-MIL, V-DIG, V-ECON, V-POL). Where audits found nothing, “No public evidence identified” is stated.
• Scale-free scoring: Impact (I) measures activity type (military application, surveillance, economic extraction, political advocacy); Magnitude (M) measures scale of involvement; Proximity (P) measures directness to Israeli/Palestinian context.
• Temporal rule: Divested or exited operations are mitigated in scoring; the audit found no OpenAI operations requiring such adjustment.
• Entity attribution: No transitive guilt — only direct relationships are scored; Microsoft’s separate IMOD contract is attributed to Microsoft, not OpenAI, though the technology flow is documented.
• Settlement operations: Dual-counting applies where operations serve both economic and political settlement functions; not applicable to OpenAI given no settlement presence.
• Vetting standard: This dossier follows the human vetting standard applied during score development — fabricated claims were rejected, divested operations discounted, wrong-entity attributions removed. The audits’ own caveats (unverified/unresolved) are carried honestly.

End Notes