ВВС США заключили пятилетний контракт на 317 млн долларов с компанией Dataminr для использования ИИ в мониторинге протестов

The U.S. Air Force awarded New York-based AI firm Dataminr a firm-fixed-price contract for A2 alerting worth $317,806,746 on August 26, 2026, locking in a company with a documented history of providing pre-demonstration surveillance alerts on U.S. protesters as the military's permanent open-source intelligence alerting provider through June 2031. It is the third escalating Air Force award to Dataminr for the same mission — following a $258.7 million contract in 2020 and a $298.8 million definitive contract modification in 2021 — and it cements the company's position as the Defense Department's primary AI-powered pipeline for turning public data streams into military force protection alerts. The award, administered by the Air Force District of Washington Contracting Directorate at Joint Base Andrews, Maryland, drew eight competing offers and obligated approximately $46.5 million in fiscal year 2026 operations and maintenance funds at signing. Work will be performed in Washington, D.C., under contract number FA7014-26-C-0026. What makes this contract institutionally significant — and what the Pentagon's framing around its "AI-first" wartime-speed strategy does not address — is that Dataminr's platform has a documented track record of being used not only for foreign threat monitoring but for alerting law enforcement to domestic, constitutionally protected protest activity. The same AI alerting infrastructure that the Air Force is now locking in for five years has, according to investigative reporting by The Intercept, supplied LAPD with protest alerts between October 2023 and April 2024 — including roughly a dozen before those demonstrations had even occurred. That capability — the ability to alert authorities to planned demonstrations before they happen — now belongs, under a firm-fixed-price subscription, to a military branch of the U.S. government. What the A2 PAI Alerting System Does The Air Force requirement, originated by Headquarters Air Force's Deputy Chief of Staff for Intelligence, Surveillance, Reconnaissance, and Cyber Effects Operations, called for a commercially available platform capable of providing "situational awareness, force protection, and indications and warnings" on global events. Dataminr's A2 Publicly Available Information Alerting system — built on its First Alert platform — delivers alerts in real time via email, web application, mobile app, and API, filtered by user-defined geographic regions and topics of interest. The technical distinction between Dataminr's approach and those of competitors like Palantir and Babel Street is important for understanding what the military is buying. Palantir's Maven Smart System, designated an official Department of War program of record in March 2026, fuses classified and unclassified data into targeting and battlespace awareness workflows. Dataminr operates exclusively in the publicly available information domain — no classified feeds, no internal government data — and its competitive advantage is speed: detecting breaking events from public sources minutes to hours before those signals coalesce into traditional intelligence reporting channels. Inside Dataminr's Multi-Modal AI Pipeline What makes Dataminr technically distinct is the architecture underpinning that speed. The company's platform processes more than 43 terabytes daily from over one million distinct sources — social media posts in 150 languages, images, video streams, audio broadcasts, and sensor data — through a multi-modal AI fusion pipeline that combines more than 50 proprietary large language models with computer vision, natural language processing, and audio analysis. Rather than relying on keyword matching to surface relevant events, Dataminr's LLMs use semantic discovery methods that allow the system to interpret context, intent, and significance — enabling it to identify a breaking threat from a social media post before the words used in that post would have triggered a conventional keyword alert. The system fuses signals across media types in a unified semantic framework: a video stream showing smoke on the horizon can be correlated with a nearby social media post and an aircraft transponder signal to produce an alert about a downed aircraft before any official reporting exists. The company calls its generative AI capability "ReGenAI," which produces contextual summaries — not just event detection flags — delivered to military users as alert text explaining what is happening, where, and why it matters to their defined area of interest. Dataminr reports detecting nearly 500,000 events daily (company-reported figure). This design is specifically engineered for the use case the Air Force is paying for: real-time situational awareness drawn entirely from unclassified public signals, delivered fast enough to give force protection decision-makers an information advantage before events escalate. The problem it cannot solve — and the contractual requirement does not ask it to solve — is the one that sits at the intersection of speed, public data, and constitutional rights. Protest Surveillance History Follows Dataminr Into Its Biggest Contract Yet Dataminr has spent six years insisting it is a "news alerting" platform, not a surveillance tool. That distinction has been repeatedly tested by investigative journalism. In July 2020, The Intercept documented that police departments received Dataminr alerts tracking Floyd protesters — monitoring that Twitter, then a Dataminr investor, had explicitly committed would not happen. Twitter told the ACLU that "the use of Twitter data for surveillance is strictly prohibited" and that Dataminr would not provide government clients with "any form of surveillance." Dataminr responded by terminating fusion center access and announcing it would not provide geospatial analysis tools to government customers. The commitment did not hold. The Intercept reported that the U.S. Marshals Service had used Dataminr to monitor pro-abortion rights protests in 2023. And in March 2025, the same outlet documented that Dataminr had alerted LAPD to 50-plus protests between October 2023 and April 2024, with approximately a dozen of those alerts issued before the demonstrations occurred — meaning the system was providing advance warning of planned protest activity to law enforcement, not merely news alerts on events already in progress. The American Civil Liberties Union's speech, privacy, and technology counsel, Jennifer Granick, described the structural concern: X (formerly Twitter) "turns around and sells Dataminr fire-hose access to users' posts, which Dataminr then passes through to the government in the form of unregulated disclosures and speculative predictions that can falsely ensnare the innocent." The University of Houston also used Dataminr to monitor pro-Palestinian demonstrations on its campus, according to The Intercept's November 2025 reporting. Dataminr has maintained that its First Alert platform identifies "breaking news events without any regard to the racial or ethnic composition of an area" and that "race, ethnicity, or any other demographic characteristic of the people posting public social media posts about events is never part of determining whether a breaking news alert is sent to First Alert clients." The company has not addressed how its platform's advance alerting on unannounced protest activity is categorically different from surveillance. Neither the new contract's publicly available documentation nor the Pentagon's AI-first strategy framework contains any language specifying how Dataminr's force protection alerting system will be prevented from being applied to domestic protest monitoring within or near U.S. military installations. Pentagon's Accelerating Bet on Commercial AI Intelligence The Dataminr award fits inside one of the most significant procurement shifts in decades. Secretary of War Pete Hegseth signed two coordinated memoranda on January 9, 2026, mandating an "AI-first" approach across all Department of War components and directing that the latest commercial AI models reach military users within 30 days of their public release. The strategy removed several prior policy constraints on AI adoption, framing speed of deployment as essential to maintaining advantage over China. That shift has been reflected in a cascade of major commercial AI contract actions. The Pentagon designated Palantir's Maven Smart System a formal program of record in March 2026, embedding it in the five-year Future Years Defense Program. In May 2026, the department opened its highest classified networks to commercial AI models from Amazon Web Services, Microsoft, Nvidia, OpenAI, Google, and xAI — what Pentagon officials described as "the final catalyst for an AI-first fighting force." The Dataminr contract represents a distinct layer of that integration: not compute infrastructure or reasoning models, but the intelligence layer — the systems that, in real time, tell military commanders what is happening in the unclassified world. The firm-fixed-price structure of the award is itself a signal: rather than a research or development contract, the government purchased a proven commercial product at scale, locking in pricing for five years. The eight competing offers indicate a market contest, though Dataminr's three prior Air Force contracts on the identical requirement gave it incumbent advantages that the GAO explicitly recognized as legitimate when Babel Street protested the 2021 award. The Brennan Center for Justice has noted the broader risk of the Pentagon's current pace: "The risks to soldiers and civilians are mounting as the Pentagon races to adopt the latest commercial advances in artificial intelligence," with the haste of deployment potentially producing "flawed weapons systems, exorbitant prices, reduced accountability, and an accelerated AI arms race." Read more: AWS, Microsoft, NVIDIA Are Giving Pentagon Access to Their AI Tech, While Anthropic Remains Out Dataminr's Expanding Defense Footprint The Air Force OSINT contract is only one dimension of Dataminr's growing defense and intelligence position. In October 2025, Dataminr announced the acquisition of ThreatConnect for $290 million in cash and equity; the deal closed in late 2025. In March 2026, Dataminr launched Dataminr for Cyber Defense, a unified product suite that fuses the company's external public-data signals with organizations' internal security telemetry — using agentic AI to automate threat investigation workflows. Dataminr's corporate advisory board now includes Jen Easterly, who served as director of the Cybersecurity and Infrastructure Security Agency from 2021 to 2025, lending institutional credibility to its government market position. The company's government intelligence footprint is not limited to the Air Force. The Defense Information Systems Agency awarded Dataminr a separate social media surveillance contract, according to reporting by researcher Jack Poulson. Dataminr has also held contracts with the Department of Homeland Security and, previously, the FBI — though its FBI social media intelligence contract was taken over by ZeroFox at the end of 2020. The escalating pattern of Air Force contract values tells its own story: $258.7 million in 2020, $298.8 million with the 2021 modification, and $317.8 million in 2026. Each award represents a larger financial commitment to the same core capability, and each award has been issued by the same contracting activity — Air Force District of Washington at Joint Base Andrews — to the same contractor. The 2026 contract extends that relationship through June 2031. Procurement Details Does Anyone Oversee What Gets Flagged? The Air Force's Dataminr requirement specifies "force protection and indications and warnings on global events" — and the system is designed around globally oriented, user-defined topical and geographic filters. But the boundary between a legitimate force protection alert (a civil disturbance near a base overseas) and an alert on constitutionally protected domestic protest activity (a demonstration near a domestic installation) is determined not by the contract's language but by how individual users configure their filters and how Dataminr's AI classifies events. No publicly available documentation of this contract — nor the Pentagon's broader AI-first strategy framework — establishes an independent oversight mechanism for what domestic activity Dataminr may alert on under the Air Force's subscription. The company's commitment to avoiding "surveillance" framing, which it has made and effectively broken at least twice since 2016, is not a legal constraint. The GAO decision denying Babel Street's protest established that Dataminr's Twitter/X firehose access was a legitimate technical differentiator in the competitive evaluation. It did not establish — and was not asked to establish — what civil liberties guardrails should govern a five-year military subscription to a platform with Dataminr's documented domestic-monitoring track record. Frequently Asked Questions Has Dataminr been used to surveil protesters in the United States? Yes, and repeatedly. Investigative reporting by The Intercept documented that Dataminr helped police departments monitor George Floyd protest locations in 2020, alerted the U.S. Marshals to abortion rights demonstrations in 2023, and provided the LAPD with more than 50 advance alerts on pro-Palestinian protests between October 2023 and April 2024 — including roughly a dozen before those demonstrations occurred. Dataminr has consistently characterized its product as "news alerting" rather than "surveillance," but the functional output in these documented cases — advance notification of planned protest activity to law enforcement — is operationally indistinguishable from protest surveillance. The company has not explained how an Air Force subscription to the same platform would be structurally constrained from producing similar alerts on activity near domestic military installations. What is the difference between Dataminr's OSINT alerting and what Palantir does for the military? Dataminr operates exclusively in the publicly available information domain — social media posts, news sources, public video and audio streams — and its competitive advantage is speed: surfacing actionable signals from public data before those signals appear in traditional intelligence reporting. Palantir's Maven Smart System, by contrast, fuses both classified and unclassified data and integrates into targeting, battlespace awareness, and command-and-control workflows. The Air Force uses Dataminr for rapid situational awareness from the unclassified open-source layer; Maven operates in the classified intelligence and targeting domain. Both are now institutional programs, but they serve structurally different functions in the intelligence pipeline. How does Dataminr's AI actually detect events in real time? The platform processes more than 43 terabytes of public data daily across one million-plus sources through a multi-modal AI pipeline combining more than 50 proprietary large language models with computer vision and audio analysis. Rather than relying on keyword matching, the LLMs use semantic discovery to interpret context and significance — enabling detection of emerging threats from posts that do not use obvious alerting language. Signals from text, images, video, and sensor data are cross-correlated in a unified semantic framework, so the system can correlate, for example, an aircraft transponder anomaly with a nearby social media image of smoke to generate an alert before any official reporting exists. Contextual summaries are then generated and delivered to users filtered by their defined geographic regions and topics of interest. What oversight exists for how the military uses Dataminr's alerts? Based on publicly available information, none that is specific to Dataminr's contract. The Air Force's requirement is written around "force protection and indications and warnings on global events," but there is no publicly documented mechanism — in the contract, the Pentagon's AI-first strategy framework, or any oversight body's published reporting — specifying how alerts on domestic protest activity would be distinguished from legitimate force protection alerts and excluded from the military's use of the platform. Congressional committees with jurisdiction over defense intelligence have not held public hearings specifically on Dataminr's DoD contracts. The company's own prior commitments to avoid providing surveillance to government clients have been documented as inadequate safeguards by investigative journalists. ⓒ 2026 TECHTIMES.com All rights reserved. Do not reproduce without permission.