Vendor Lock-In And Operational Continuity Risk From Contractual Or Technical Constraints

Issue 72 Edition 2026-03-13 8 min read

General

Sources: 1 • Confidence: Medium • Updated: 2026-03-14 12:23

Key takeaways

Emil Michael claims that reviewing prior-administration AI contracts, he found dozens of usage restrictions, including prohibitions on using AI for planning or executing operations that could lead to kinetic strikes.
Emil Michael claims China is stealing U.S. AI models and removing guardrails, implying U.S. restrictions could create asymmetric disadvantage if adversaries deploy the same models without constraints.
Upon taking office, Emil Michael found the DoD had 14 critical technology priority areas that had not materially changed in nearly a decade and were too vague to drive action.
Emil Michael claims prime contractors' main advantage over startups is scaled manufacturing and production capability rather than inventiveness, and he expects startups to build that capability over the next one to two years.
Emil Michael says the DoD is shifting away from complex RFP requirement checklists and cost-plus development toward simpler outcome-based requirements with multiple competing approaches and firm fixed-price contracting.

Emil Michael claims that reviewing prior-administration AI contracts, he found dozens of usage restrictions, including prohibitions on using AI for planning or executing operations that could lead to kinetic strikes.
Emil Michael asserts that commercial AI models were already embedded in highly sensitive U.S. military commands including CENTCOM, INDOPACOM, and SOUTHCOM under terms that created vendor lock-in.
Emil Michael says the DoD must engage enough AI companies so it is never single-threaded on one vendor again for critical capabilities.
Emil Michael contends that if model terms were enforced by shutdown logic, a vendor could theoretically cause AI tooling to stop mid-operation upon a perceived terms violation, creating operational risk to personnel.
Emil Michael says a senior executive at a primary AI vendor questioned whether the vendor's software was used in the 'Maduro raid' and implied they might not want it used for such operations.

Emil Michael claims China is stealing U.S. AI models and removing guardrails, implying U.S. restrictions could create asymmetric disadvantage if adversaries deploy the same models without constraints.
Emil Michael argues that because AI is becoming a general substrate, the U.S. military must be able to use it for lawful purposes without being constrained by a vendor's internal values document.
Emil Michael contends that decisions balancing civil liberties and national security should be set through Congress and regulation rather than by unilateral choices of AI company leadership.
Emil Michael says adversaries use AI partly to reduce reliance on human decision-making due to lower internal trust, whereas the U.S. aims to use AI to enhance human decision-making.

Upon taking office, Emil Michael found the DoD had 14 critical technology priority areas that had not materially changed in nearly a decade and were too vague to drive action.
Emil Michael reduced the DoD's critical technology priorities from 14 to 6, made applied AI the top priority, and moved the Chief Digital and AI Office into his organization.
Emil Michael frames DoD AI use cases into three buckets: enterprise efficiency, intelligence analysis from large data repositories, and warfighting logistics/simulation.

Emil Michael claims prime contractors' main advantage over startups is scaled manufacturing and production capability rather than inventiveness, and he expects startups to build that capability over the next one to two years.
Emil Michael attributes slow DoD acquisition speed to post-Cold War defense-industry consolidation that followed a Pentagon signal to slow growth, which he says left the U.S. behind as China began a major military buildup in the mid-2000s.
Emil Michael describes the frontier AI landscape as effectively four frontier companies competing over roughly a thousand highly valuable researchers who are frequently traded among firms.

Emil Michael says the DoD is shifting away from complex RFP requirement checklists and cost-plus development toward simpler outcome-based requirements with multiple competing approaches and firm fixed-price contracting.
Emil Michael says DoD culture tends to avoid saying no, and he is pushing for faster yeses and faster nos so startups can reliably decide whether to proceed or move on.

What were the exact six critical technology priorities after the reduction, and what measurable criteria were used to make them actionable?
How is DoD AI “usage” defined in the reported jump, what telemetry or reporting system measured it, and what qualifies as a user?
Which contracts contained the cited operational-use restrictions, and what exact clauses prohibited planning or execution related to kinetic strikes?
What specific terms or technical integration patterns created the claimed vendor lock-in within sensitive combatant commands, and what would make those systems portable?
Do any deployed systems include technical kill-switches or suspension mechanisms tied to terms-of-service enforcement, and what continuity-of-operations protections exist?

Defense buyers may favor multi-vendor, portable AI architectures to reduce vendor lock-in and continuity-of-operations risk from contractual or technical constraints, potentially shifting spend toward interoperable tooling and away from single-provider embedded deployments.
Acquisition reform toward outcome-based requirements and firm fixed-price competition could advantage vendors able to deliver measurable results quickly, while reducing appeal of checklist-driven proposals and cost-plus development models.
Greater emphasis on model security and control of guardrails, driven by concerns about model theft and asymmetric constraints, could increase demand for secure deployment, monitoring, and governance features aligned to lawful authority rather than vendor terms.

Solicitations and awards explicitly require multi-vendor capability, portability, data and model egress, or continuity protections against service suspension during operations.
Procurement language shifts toward outcome metrics, multiple parallel prototypes, faster downselects, and firm fixed-price structures, with reduced reliance on large requirement checklists.
DoD publishes a reduced set of actionable critical technology priorities with measurable criteria, and governance consolidation results in standardized security and policy requirements for AI deployments.

DoD continues to award single-vendor embedded AI deployments without portability, or accepts terms allowing suspension tied to vendor enforcement, indicating lock-in concerns are not translating into procurement practice.
Outcome-based and firm fixed-price contracting does not materialize at scale, with continued dominance of cost-plus development and checklist-heavy RFPs.
No visible tightening of model security or governance requirements, and limited action on theft and guardrail control concerns beyond rhetoric.