Understanding DoD's Authority
This guide explores the Department of Defense's (DoD) power to label contractors as supply chain risks under the Federal Acquisition Supply Chain Security Act (FASCSA), focusing on the D.C. Circuit Court's September 2026 ruling in the Anthropic case. This decision sets a precedent for how product functionality restrictions might be interpreted. If your team is implementing technical safeguards, compliance controls, or usage limitations, this ruling could impact how DoD views those design choices.
This guide does not cover First Amendment challenges to FASCSA designations or the separate California district court ruling in favor of Anthropic.
Key Concepts
Supply Chain Risk: Under FASCSA, this risk involves the potential for an adversary to sabotage or subvert a covered article's design or operation, affecting its function or security.
Manipulation: The D.C. Circuit determined that manipulation doesn't require malicious intent. It involves any deliberate configuration of product functionality, whether for protection or otherwise.
Covered Article: These are products or services subject to FASCSA review, including software, cloud services, and AI models under federal contracts.
GenAI.mil Platform: The DoD's generative AI deployment environment mentioned in the Anthropic contract dispute.
FASCSA Framework
FASCSA allows the Secretary of Defense to exclude sources or articles from procurement if they pose a supply chain risk. The statute doesn't require:
- Proof of adversarial intent
- Evidence of foreign influence
- Demonstration of actual harm
- Procedural review beyond internal DoD determination
The Secretary only needs to determine that a supply chain risk exists.
Implications of the D.C. Circuit Ruling
In the Anthropic case, the court ruled that:
Product functionality restrictions are manipulation: Training an AI model to refuse tasks shows the ability to control its actions.
Intentions don't exempt manipulation: The court found Anthropic's motivations irrelevant to the statutory analysis.
Operational risk is enough: The potential for models to shut down during operations presents national security concerns.
The court rejected the need to prove a causal link between a contractor's statements and the supply chain designation.
Contract Value and Stakes
Anthropic's contract was valued at up to $200 million over two years. The ruling shows that even significant contracts don't limit DoD's FASCSA authority.
Guidance for Contractors
Building Products for DoD
Document Design Rationale: Clearly separate design rationale from operational constraints. Label security controls and functionality restrictions as compliance obligations, not preferences.
Assess Task Refusal Capabilities: Evaluate if your product's refusal logic could be seen as a supply chain risk.
Consider Operational Continuity: Document how safety mechanisms interact with mission-critical operations and whether you can provide override capabilities.
Current Contract Holders
Review Technical Specifications: Understand that DoD may reinterpret usage restrictions as supply chain risks.
Prepare for Litigation: Venue selection matters if facing FASCSA designation. The D.C. Circuit applied a deferential review to DoD's determination.
Monitor Public Statements: Public positions on safety or ethics won't protect you from FASCSA action.
Common Pitfalls
Assuming Compliance Exempts You: Implementing controls like NIST SP 800-171 doesn't shield you from designation.
Believing Contractual Performance Protects You: Existing contracts don't limit DoD's authority.
Treating Safety Features as Technical Decisions: Your safety analysis doesn't override DoD's risk assessment.
Expecting Narrow Application: The ruling's precedent extends beyond AI models.
Assuming Other Companies Will Ignore the Precedent: The ruling may deter companies from implementing similar safeguards.
Quick Reference Table
| Element | D.C. Circuit Standard | Contractor Implication |
|---|---|---|
| Intent requirement | None, based on actions, not intentions | Noble intentions don't exempt you from designation |
| Manipulation definition | "To arrange, operate, or control skillfully" | Any deliberate configuration qualifies |
| Proof of harm | Not required | No need to show actual compromise |
| Operational risk | Service disruption potential suffices | Products with refusal logic face scrutiny |
| First Amendment protection | Rejected, no causal link needed | Public statements won't shield you |
| Contract protection | None, existing contracts don't limit authority | Performance history irrelevant |
| Review standard | Deferential to Secretary of Defense | Challenging designation in D.C. Circuit is difficult |
| Appeal options | Circuit split creates venue questions | Forum selection matters |
Bottom line: The D.C. Circuit's Anthropic ruling interprets FASCSA broadly, covering any product functionality restriction. Document how features serve compliance obligations rather than discretionary choices. Don't assume your contract, security posture, or public commitment to responsible AI protects you from supply chain risk designation. DoD's operational risk assessment takes precedence over your technical safety analysis.
