Cordova 2028 Releases AI Security Doctrine: "America Must Not Limit Its Own People"

Independent 2028 presidential candidate Vincent Dean Cordova, Jr. today published two documents — an open letter to the American people, "America Must Not Limit Its Own People," and a policy brief, "Why Capability Denial Fails: An AI Security Doctrine Built on Collaboration, Not Containment." Both respond to the June 2026 federal export-control directive that, according to Anthropic, required the company to suspend access to its advanced Fable 5 and Mythos 5 models for foreign nationals — a directive whose scope led the company to disable the models for all users while it worked to comply.

Cordova's central argument is that the restriction targeted the wrong variable. The order limited who could use a tool and where they were located, rather than the specific capability, the specific use, and whether the limit could actually be enforced. Because the capability at issue — having an AI read code to find software flaws — is knowledge that can be copied and rebuilt, Cordova argues that restricting it at one domestic builder slows lawful, visible Americans while leaving actors who ignore the rules untouched.

“We cannot claim to protect America by making Americans less capable. That is not national security — that is national disadvantage. America's answer to technological risk cannot be to reduce the capability of its own people. America's answer must be readiness.”

The brief draws a deliberate line between capabilities that can and cannot be contained. The decisive variable is whether a capability is gated by diffusible knowledge or by a physical chokepoint.

Diffusible capabilities — code, algorithms, model techniques, prompts, workflows, and knowledge — cannot be durably contained. The marginal cost of copying approaches zero. When a government restricts a diffusible capability at one source, it does not remove the capability from the world. It removes it from the actor it can see and regulate, while actors it cannot regulate proceed unaffected.

Restrictions on diffusible knowledge mostly penalize those who comply, Cordova argues, while limits tied to genuine physical chokepoints — such as advanced manufacturing inputs — may be enforceable and appropriate. The document explicitly concedes that some controls are legitimate, a distinction Cordova says is necessary for any honest security doctrine.

This is not a new lesson. In the 1990s, the United States treated strong encryption as a controlled export. The underlying mathematics was published and rebuilt regardless, and the rules were later relaxed. The pattern is the same here: a control that can be routed around buys little durable safety and imposes real cost on the people who obey it.

Cordova frames the core challenge as a contest between defense and offense rather than a problem that restriction can solve. AI is already being used to find vulnerabilities in software. That is not a future risk to be prevented; it is a present reality to be managed. Safety in this domain is a race condition — a continuous contest between defensive use and offensive use — and the policy question is which side our choices accelerate.

Restriction loses that race by construction. It slows the defenders who operate in the open and submit to oversight, while leaving untouched the actors who never asked permission. The same capability that can probe a system for weaknesses is the capability that lets defenders find and close those weaknesses first.

He points to publicly reported results from Anthropic's Project Glasswing, in which roughly 50 partner organizations used a restricted model to identify more than 10,000 high- or critical-severity software vulnerabilities across important software systems. "Those numbers are not an argument for carelessness," the brief reads. "They are an argument for scaling defense faster than offense."

“The question is not whether we can stop that. We cannot. The question is whether our defenders stay ahead of those who would do harm — and denial slows the defenders we can see while doing little to the actors we cannot.”

The brief closes with four pillars of a collaboration-based AI security doctrine that Cordova says takes risk seriously without surrendering the benefit:

1. Tier by capability, not by identity. Most of what makes these tools transformative carries no catastrophic risk and should be widely available. Restrict the genuinely dangerous slice at the capability tier for everyone — with transparent standards and vetting — rather than using nationality as a substitute for technical judgment.

2. Universal risk literacy. Broad access should come with broad understanding. Train the public to recognize accident-shaped dangers: data exposure, over-trust in outputs, false certainty, and weak review practices.

3. Defense and resilience investment. Because dangerous knowledge will diffuse regardless of any one restriction, the responsible assumption is that it gets out — and the work is to harden the targets. Public investment in defensive security and secure-by-design software is where a sovereign government's effort belongs.

4. Narrow, enforceable limits at real chokepoints. Reserve hard restriction for cases where a genuine physical bottleneck exists, review those limits regularly, and refuse the comforting illusion that fencing a diffusible capability at a domestic builder makes anyone safer.

The brief closes with a proposed governing standard for any future restriction on emerging technology. Cordova would require the government to identify the specific capability being restricted, show a concrete risk beyond ordinary dual-use concern, apply the standard evenly across equivalent systems, demonstrate that the limit is technically enforceable, provide a response and restoration process for the affected party, and use the narrowest restriction capable of addressing the risk.

"This standard does not eliminate government authority," the brief states. "It disciplines it." The brief is careful to note what it does not claim — that any company was deliberately persecuted, that the directive was driven by commercial or political motive, or that all AI restrictions are illegitimate. Those questions are contested or unproven. The case rests on the public record and the structure of the technology itself.

“Capability must belong to the people, not only to capital. The future cannot be reserved for those with contracts, lobbyists, and institutional permission. We must advance Americans — fully, fairly, and first.”

The open letter, "America Must Not Limit Its Own People," addresses the broader pattern: public protections being weakened, economic power concentrating, and at the same time the tools that could help Americans compete, build, defend, learn, and rise being restricted or gated. Cordova frames readiness — not denial — as the correct national posture: universal AI literacy, small-business access, worker protection, educational access, defensive strength, and public oversight of private-equity capture.

The companion policy brief, "Why Capability Denial Fails," provides the technical and structural argument, draws on the public record regarding the June 2026 federal directive, and presents the governing standard for future AI restrictions. Both documents are available for download at the links below.

• ↓Policy Brief PDF — Why Capability Denial Fails: An AI Security Doctrine Built on Collaboration, Not Containment
• ↓Open Letter PDF — America Must Not Limit Its Own People
• ↓Press Release PDF — Cordova 2028 Releases AI Security Doctrine

Vincent Dean Cordova, Jr. is an independent candidate for President of the United States in 2028, based in Manteca, California. Running under the banner Moving Forward Together, his platform centers on broad public access to opportunity, accountability for concentrated power, worker and community ownership, and the principle that technology should expand human dignity rather than become another system of control.