What project MSL-05 buys
The Artificial Intelligence and Human-Machine Symbiosis project develops technologies to enable machines to function not only as tools that facilitate human action but also as trustworthy partners to human operators. Of particular interest are systems that can understand human language, extract information, and reliably categorize content contained in diverse media; answer questions, reach conclusions, and propose explanations; and learn, reason, and apply knowledge gained through experience to respond intelligently to new and unforeseen events. Enabling computing systems with such human-like intelligence matters because the tempo of military operations in emerging domains can exceed that at which unaided humans can orient, understand, and act. The technologies developed in this project will enable warfighters to make better decisions in complex, time-critical, battlefield environments with greater efficiency and confidence; and help autonomous systems and intelligent computational agents to safely and reliably perform critical missions in contested and adversarial physical and virtual environments. Prior to FY 2026, efforts in this Project were funded in PE 0602303E, Project IT-04.
Project MSL-05 funding, FY2025–FY2031
Prior years are actuals, the budget year is the request, and the outyears are the FYDP plan. Estimate types are colored and never summed into one figure. Projects carry the full five-year plan; the activities inside them stop at the budget year.
| Fiscal Year | Estimate Type | Amount ($M) |
|---|---|---|
| FY2025 | Actual | 0.0 |
| FY2026 | Enacted | 120.7 |
| FY2027 | Request | 139.0 |
| FY2028 | Outyear | 202.4 |
| FY2029 | Outyear | 237.2 |
| FY2030 | Outyear | 248.4 |
| FY2031 | Outyear | 251.4 |
14 accomplishments / planned programs
The R-2A exhibit — the only level of the budget that describes work that has not happened yet. Activities carry the prior, current and budget year only, no five-year plan. 5 of them describe FY2027 work in enough detail to have their own page; the rest are shown here in full. Coverage is partial across the corpus, so count activities, never total them.
- Develop formal approaches for software adaptation and system behavior prediction in the field. - Create physics-based models of perception for high reliability AI/ML-based sensing in autonomous systems. - Integrate distributed control, game-theoretic incentives, and strategic reasoning to enable robust, controllable, and adaptive…
Read the FY2027 plan →- Enable enhanced collaborative reasoning by integrated human-machine warfighting formations. - Automatically generate high-fidelity digital representations and behavioral models of adversary groups. - Develop AI for automated reverse engineering during active cyber warfare. - Develop AI that continually improves its model as it operates…
Read the FY2027 plan →- Refine AI risk analysis techniques and tools based on the results of initial operational exercises and red team experiences. - Quantify the effectiveness of alternative robustness enhancements and TTPs across diverse use cases. - Conduct high-fidelity operational and red team exercises targeted to high-priority DoW AI-robustness use…
Read the FY2027 plan →- Conduct the first at-scale evaluation of genAI assessment methods for compositions of classes of problems. - Integrate mathematical guarantees into assessment methods at scale. - Formalize best practices for classes of problems and compositions of classes to facilitate adoption by the genAI user community.
Read the FY2027 plan →- Improve TTP detection performance through adaptive generative approaches and demonstrate capability to anticipate and extract novel money laundering TTPs from heterogeneous FININT and transaction data. - Collaborate with AML stakeholders from across USG and the private sector to evaluate techniques and tools in operationally relevant…
Read the FY2027 plan →FY2027 planned work - Evaluate R&D efforts addressing critical technical challenges associated with the use of frontier AI for national security and initiate new efforts as appropriate. - Launch additional collaborative R&D efforts that address critical technical challenges associated with the use of frontier AI for national security.
FY2026 to FY2027 change The FY 2027 increase reflects ramping up of R&D efforts that address critical challenges associated with the use of frontier AI for national security.
FY2026 plans — current year - Complete an acquisition strategy and management plan for a joint government-university-industry project that features R&D efforts that solve critical challenges associated with the use of frontier AI for national security. - Formalize agreements with government and industry partners to establish processes for initiation, approval, and review of R&D efforts focused on the use of frontier AI for national security.
FY2027 planned work - Extend core capabilities to include the estimation of risk and reward structures associated with belief states. - Transition Kallisti technologies.
FY2026 to FY2027 change The FY 2027 decrease reflects the shift from developing models that estimate actor comprehension to developing models that estimate actor strategies.
FY2026 plans — current year - Develop relevant context of beliefs and system dynamics to constrain belief space. - Actively manage multiple reference frames using less data and computational complexity. - Use novel methods in moment matching to improve cross-entropy of observations.
FY2027 planned work - Attract private capital to support efficient operation of the forum and operationally relevant development of geoeconomic theory and playbooks. - Develop geoeconomic playbooks derived from geoeconomic research.
FY2026 to FY2027 change The FY 2027 decrease reflects the transition of the forum from DARPA-provided funding to private capital.
FY2026 plans — current year - Employ precisely described national security outcomes to iteratively design and rigorously evaluate geoeconomic mechanisms. - Hold symposia with USG and industry to precisely define national security objectives and their related economic incentives to inform geoeconomic research objectives. - Generate theoretically valid economic mechanisms that concurrently produce positive national security outcomes and U.S. economic benefits. - Test the empirical validity of proposed economic mechanisms to ensure they meet desired national security outcomes while beneficially incentivizing U.S. economic activity.
FY2027 planned work - Transition techniques to Department of War program offices that can benefit from enhanced capabilities to simulate autonomy.
FY2026 to FY2027 change The FY 2027 decrease reflects a shift from development to transition of techniques.
FY2026 plans — current year - Develop methods for measuring the sim2sim and sim2real gaps, as well as robustness to these gaps. - Demonstrate a capability to transfer autonomy from abstract simulations to live platforms for scenarios of interest to potential transition partners. - Conduct a sim2real competition in collaboration with private sector stakeholders to evaluate the robustness of simulated autonomy in a real-world platform subject to large sim2real gaps.
FY2027 planned work - Collaborate with transition partners to leverage feasibility assessment tools in operational settings.
FY2026 to FY2027 change The FY 2027 decrease reflects a shift from tool evaluation to transition.
FY2026 plans — current year - Collaborate with Department of War and Intelligence Community transition partners to evaluate the utility of the feasibility assessments and to refine techniques.
FY2026 to FY2027 change The FY 2027 decrease reflects program completion.
FY2026 plans — current year - Complete final version of structural price prediction models and extend to additional minerals. - Complete final supply and demand forecast models and extend to additional minerals. - Conduct evaluation of completed structural price prediction and supply and demand forecast models. - Conduct research surrounding application of transparent structural price and supply/demand forecast methodology to additional areas. - Transition final models to government and commercial partners.
FY2026 to FY2027 change The FY 2027 decrease reflects program completion.
FY2026 plans — current year - Demonstrate cooperative operation of multiple controller instances in a single, complex physical system. - Collaborate with transition partners to demonstrate safe operation and functionality on real-world, operational platforms despite damage or malfunction. - Formalize mathematical foundations of the recovery and control techniques for adoption by industry and academia for robust control system design.
FY2026 to FY2027 change The FY 2027 decrease reflects program completion.
FY2026 plans — current year - Finalize AI neurosymbolic datasets and benchmarks and conduct demonstrations on selected use cases. - Transition methods to DoW.
FY2026 to FY2027 change The FY 2027 decrease reflects program completion.
FY2026 plans — current year - Organize intellectual property generated for use by the U.S. Government and Department of War.