US DoD Chief Digital and AI Officer Craig Martell says the agency wants to use AI safely and urges companies to be more transparent about their AI software
- Defense Department is holding symposium to discuss AI — Official says agency wants to use the algorithms safely X: @alfred_lanning1 , @katrinamanson , and @lynnmdoan X: Alfred Lanning / @alfred_lanning1 : The Defense Department's top artificial intelligence official said the agency needs to know more about #AI tools before it fully commits to using the technology and urged developers to be more transparent. https://www.bloomberg.com/... Katrina Manson / @katrinamanson : Pentagon's AI chief says DOD has no idea how generative AI models are structured or what data has been used. “We're just getting the end result of the model-building — that's not sufficient,” Craig Martell tells me. https://www.bloomberg.com/... Lynn Doan / @lynnmdoan : The pentagon has an AI chief.
Context & Ripple Effects
The Pentagon had already set ethical principles for its use of AI before creating a dedicated digital and AI leadership role, filled by Martell in 2022 through the department's first chief digital and AI officer appointment.
Martell's comments make model structure and training-data visibility a practical condition of the Defense Department's safe-use agenda, rather than an abstract governance preference.
First-order effects
- AI developers seeking Defense Department use face a clearer demand to explain how their generative models are built and trained.
- The department is likely to constrain or slow commitments where it cannot assess a model's provenance and structure sufficiently for safe use.
Second-order effects
- Transparency becomes a procurement differentiator: vendors able to supply usable model documentation gain an advantage over providers that treat core development details as inaccessible.
- The department's AI symposium and internal review processes can shift adoption toward tools whose risks can be evaluated, not simply those with the broadest capabilities.
Third-order effects
- If repeated across government buying, AI suppliers may increasingly need to build products and disclosure practices suited to high-assurance public-sector deployment.
- The longer-term tension is between commercial model opacity and institutional accountability; procurement requirements could become a key mechanism for resolving it, though the eventual standard remains uncertain.
The trend: This is one data point in the rise of operational AI governance, where access to consequential deployments depends on demonstrable safety and traceability.