Pentagon gives Casepoint $98.8M BPA for classified legal AI
The award puts AI into the legal records machinery where defensibility matters more than demo-room cleverness.
TL;DR
DefenseScoop reports that the Pentagon awarded Casepoint an exclusive blanket purchase agreement worth up to $98.8 million for AI-enabled eDiscovery software, support and training. The work covers the DOD Office of General Counsel, DISA OGC, the Defense Legal Services Agency and 28 DLSA OGC offices. Casepoint says its platform does not rely on outside generative AI models, which is useful, but still leaves the hard question: how classified legal workflows will prove the AI stayed inside the evidentiary lines.
The Pentagon’s legal shops are buying AI for the least glamorous part of the enterprise and, operationally, one of the more consequential ones: finding, reviewing and producing sensitive records under deadlines that do not care how many repositories the data lives in. According to DefenseScoop, Casepoint announced an exclusive blanket purchase agreement worth up to $98.8 million to provide AI-enabled eDiscovery software as a service, support services and training for the DOD Office of General Counsel, the Defense Information Systems Agency Office of General Counsel, the Defense Legal Services Agency and 28 DLSA OGC offices.
That is not battlefield AI. It is FOIA response work, mandatory declassification review, litigation support, investigations, Inspector General and Government Accountability Office audits, congressional inquiries and the rest of the legal operations backlog that tends to become a compliance problem only after the clock has already started. Casepoint’s chief executive told DefenseScoop the platform does not rely on generative AI models from other companies and framed AI governance as a business, legal, security, compliance and product responsibility.
For practitioners, the interesting piece is less the word AI than the word defensible. Classified legal operations do not get much value from a tool that can summarize quickly but cannot show how it handled access, provenance, retention, review decisions and production boundaries. A vendor-run eDiscovery platform can reduce manual drag across fragmented repositories. It also has to make its automation legible to lawyers, records officials, security staff and auditors who will be asked later why a document was found, withheld, produced or missed.
Published ·Deep Fathom