The Department of War’s Chief Digital and Artificial Intelligence Office increased its enterprise agreement with Scale AI from $100 million to $500 million today. This five-fold increase follows the rapid uptake of the Production Other Transaction Authority agreement originally awarded in September 2025. Demand across the Department exceeded the original ceiling, forcing the expansion to support computer vision, generative AI decision-support, and data operations. The CDAO Production OTA allows any Department of War component to route funding to the centralized contracting authority and initiate its own Project Agreement for any Scale product, service, or capability without requiring a new competitive solicitation. This flexibility facilitates the ability of teams to move at mission speed.
The expansion helps the Department of War accelerate the adoption of data and AI capabilities across the joint force. This partnership supports the Department of War AI Strategy, which focuses on building a robust data foundation and creating an agile, iterative pipeline for AI deployment at the tactical edge. Scale AI provides the infrastructure to transform raw, sensitive data into AI-ready assets through its Data Engine. This development allows for the deployment of state-of-the-art models in months instead of years. The infrastructure decisions made today will shape U.S. military operational capabilities for the decade ahead.
Technical capabilities for the joint force
The Department of War can access a full suite of AI capabilities through this expanded agreement. Scale Data Engine provides a machine learning operations platform to build, test, and deploy computer vision models. This platform relies on expert-labeled, AI-ready data from the Scale AI Center in St. Louis. Annotators at this center train specifically for defense-domain tasks, including labeling EO/IR and SAR imagery for autonomy programs and generating ground truth for aided target recognition and sensor fusion. Scale GenAI Platform provides government builders a way to fine-tune, test, evaluate, and deploy generative AI models safely on classified networks.
Scale Donovan serves as a generative AI decision-making platform for defense and intelligence operators. This platform turns massive amounts of unstructured data into actionable insights and decision support at mission speed. Scale also provides engineering capability development sprints, which are discrete, time-bound, deliverable-based data and software efforts tailored to specific mission requirements. These capabilities work across NIPR, SIPR, and JWICS networks. The Scale Data Engine has already supported the generation of mission-specific, expert-validated data for various components.
The use cases for these tools span the Department’s highest priority missions, including autonomy and homeland defense. Scale Donovan has become popular with combatant commands for processing battlefield intelligence. Scale GenAI Platform helps build agentic solutions that automate complex workflows. Scale works with a diverse coalition of industry and government partners, including defense innovation ecosystem members like the DIU and various Combatant Commands.
| Capability | Primary Function | Deployment Environment |
|---|---|---|
| Scale Data Engine | Build, test, and deploy computer vision models using expert-labeled data | NIPR, SIPR, JWICS |
| Scale GenAI Platform (SGP) | Fine-tune, test, evaluate, and deploy generative AI models | NIPR, SIPR, JWICS |
| Scale Donovan | Transform unstructured data into actionable insights and decision support | NIPR, SIPR, JWICS |
| Engineering Sprints | Bespoke software and data for specific mission requirements | NIPR, SIPR, JWICS |
Regulatory volatility and the Anthropic impasse
The regulatory landscape for AI contractors is shifting rapidly. In February 2026, the Department of War and Anthropic reached an impasse over the lawful use of AI models. Anthropic refused to permit its models to be used for mass domestic surveillance or in fully autonomous weapons systems that identify and engage targets without human intervention. In response, Secretary of War Pete Hegseth directed the Department to designate Anthropic a supply chain risk. This designation follows a February 27, 2026, directive from President Donald Trump to halt the use of Anthropic technology.
The Department of War maintains that technology must be freely usable within a lawful use framework. Secretary Hegseth argued that the Department should not be restricted by a supplier’s internal usage rules. This tension highlights the difficulty of navigating the growing set of obligations under Executive Order 14409 and National Security Presidential Memorandum 11. These documents overlay the CMMC program and the Department of War policy on autonomy in weapons systems, DoDD 3000.09. The National Security Presidential Memorandum 11 specifically directs agencies to terminate contracts with companies that demonstrate a pattern of conduct inconsistent with its accountability pillars.
OpenAI signed a different agreement with the Pentagon on February 28, 2026, to integrate its models into the Department’s classified network. OpenAI’s contract includes three binding red lines to ensure safe and secure deployment. The agreement prohibits the use of OpenAI technology for mass domestic surveillance and prohibits the technology from directing autonomous weapon systems. The contract also prohibits the use of OpenAI technology in high-risk automated decision systems, such as social credit-like systems. OpenAI’s deployment remains cloud-only, which prevents the distribution of models to edge devices. This architecture ensures that the models cannot be directly integrated with weapon systems, sensors, or operational hardware.
National security perimeters in the NDAA
Congress provided specific authorities in the National Defense Authorization Act for Fiscal Year 2026 to manage AI security. Section 1512 requires the Secretary of War to establish a Department-wide cybersecurity and governance policy for AI/ML systems. This policy must address AI-specific threats, security best practices, monitoring, and testing standards. Section 1513 requires the Department to develop physical and cybersecurity standards for covered AI/ML technologies. These standards must address procurement risk, timelines, and the integration of the Cybersecurity Maturity Model Certification. Covered technology includes acquired systems and lifecycle components like trained parameters, development methods, and algorithms.
Section 1532 of the NDAA bans the acquisition or use of covered AI developed by certain entities. It also required the removal of covered AI from Department of War systems by January 17, 2026. These requirements move the Department and its contractors toward affirmative AI-specific requirements focused on policy, the defense supply chain, and protection from adversary-linked technology. Contractors must manage these requirements as part of their broader cybersecurity and engineering workstreams. The Department of War also issued a guidance to update DoDD 3000.09 in late 2026.
You probably know that the Pentagon is moving toward autonomy, but the speed of this specific contract expansion is different. The Department of War is buying autonomous systems faster than it is writing the rules to govern them. Contractors must comply with a patchwork of authorities while remaining nimble enough to prepare for regular regulatory changes. The Department of War also released an Intellectual Property Guidebook in April 2025 to address the gap in how legacy DFARS rules handle evolving AI models.
Labor disputes and psychological risks
Scale AI faces significant legal and regulatory scrutiny regarding its treatment of data-labeling contractors. The Department of Labor is investigating the company for compliance with the Fair Labor Standards Act. This investigation focuses on fair pay standards and working conditions for the company’s tens of thousands of contributors. The probe began almost a year ago during the previous administration but only recently became public. Scale AI disputes these allegations and maintains that it complies with all labor laws.
In January 2025, several contractors filed a complaint alleging workplace psychological injury. The plaintiffs, known as taskers, claim they developed PTSD, depression, and anxiety due to repeated exposure to traumatic content. This content included information regarding suicidal ideation, predation, child sexual assault, and other highly violent topics. The lawsuit also alleges moral injury and institutional betrayal. The plaintiffs contend they received insufficient warning, support, or workplace safeguards while coding this information.
The litigation in the Northern District of California examines how workers experience conduct that could be construed as unlawful during AI training. If the plaintiffs succeed, the case could introduce changes to the data-labeling industry, such as more comprehensive disclosures and mental health resources. Scale AI has faced previous scrutiny over labor practices. In 2023, reports mentioned overseas contributors complaining about low pay and impossible demands. The company maintains that it provides flexible work opportunities and that most payment inquiries resolve within three days.
Legal challenges of autonomous weapons
The development of autonomous weapons systems introduces complex challenges to international human rights law. These systems select and engage targets based on sensor processing rather than human inputs. This capability challenges the principles of distinction and proportionality. The principle of distinction forbids attacks that cannot reliably distinguish between military objectives and civilians. Because AI relies on algorithms, it may struggle to identify subtle cues of human behavior.
The principle of proportionality requires that expected civilian harm is not excessive relative to the military advantage. AI systems rely on preset metrics, which makes it difficult to weigh real-world context and human lives in the same way a person does. The Martens Clause also provides a safeguard, requiring that warfare practices comply with the principles of humanity and the dictates of public conscience. Many organizations advocate for meaningful human control to ensure that humans maintain oversight over target selection and engagement.
The lack of human judgment in autonomous systems creates an accountability gap. When a machine selects and attacks a target independently, current legal rules struggle to attribute responsibility for the resulting harm. It becomes difficult to prove intent or negligence among developers, military leaders, or deployers. This difficulty makes it problematic to hold parties responsible for infractions committed by autonomous systems. The ability of AI to adapt its own programming independently also makes it difficult to ensure that systems remain predictable and controllable.
Market competition and the path to scale
Scale AI holds a strong position in the AI infrastructure market, but it faces intense competition from other defense tech firms. The company reached a valuation of nearly $14 billion after a funding round led by Accel. Meta Platforms owns 49 percent of the company following a $15 billion investment. Scale AI reported revenue of $870 million in 2024 and expects to reach $1.4 billion by the end of 2024. This growth comes as the company moves from small-scale pilots to enterprise-wide deployment.
Anduril is a major competitor in the defense space, with a valuation of approximately $61 billion and 2025 revenue of $2.2 billion. Anduril’s strength lies in its ability to combine software with physical hardware, such as aircraft, missiles, and underwater vehicles. Anduril uses its Lattice platform as an integration layer to connect various sensors and autonomous products. This allows the company to enter a military customer through one system and expand into more workflows.
Scale AI follows a different path by focusing on the data and software infrastructure that makes models work. While companies like Anduril build the machines, Scale AI builds the data foundations and the training environments. This distinction is important as the Department of War seeks to scale AI capabilities across the joint force. Some analysts compare Anduril to Palantir because of its ability to integrate into command-and-control architectures. Scale AI remains a key player in providing the underlying data that all such platforms require.
| Company | 2024/2025 Revenue | Reported Valuation | Primary Focus |
|---|---|---|---|
| Scale AI | $870 million (2024) | ~$14 billion | AI Data & Infrastructure |
| Anduril | $2.2 billion (2025) | ~$61 billion | Autonomous Hardware & Software |
| Palantir | $4.5 billion (Recent) | Not specified | Data Integration & Analytics |
The Department of War’s decision to expand its agreement with Scale AI shows a clear preference for integrated, data-driven solutions. The ability to bypass traditional, multi-year acquisition cycles through the CDAO OTA provides a significant advantage for contractors. Scale AI provides the essential data services that enable the Department to operationalize its AI ambitions. However, the company must still manage the legal risks associated with its labor practices and the ethical implications of its defense products. Will the Department of War successfully balance the need for speed with the requirement for responsible AI?
