AI & Labor Policy · Complete Framework
AI works for people. People are not replaced by AI. A complete operational framework — mandating human-AI collaboration, closing corporate and offshore loopholes, banning foreign government control of American AI infrastructure, and protecting animals and ecosystems from the same extractive mandate threatening the American workforce.
The Governing Principle
“AI works for people. People are not replaced by AI.”
This is the complete operational framework underlying the Cordova administration’s Executive Order and the companion American Worker Protection and Human-AI Collaboration Act. It addresses three converging threats simultaneously: AI-driven workforce displacement directed by extractive financial structures, offshore and structural circumvention loopholes, and foreign government acquisition of control points in American critical infrastructure. And it extends the same principle to the natural world — because the extractive mandate that threatens American workers also threatens every living thing it touches.
771,000+
Americans experiencing homelessness
39M
Americans living in poverty
300M
Jobs at risk globally from generative AI (Goldman Sachs)
$500K
Maximum civil penalty per affected worker per violation
Policy Framework
01 · The Governing Principle
Every provision of this policy flows from one sentence. AI deployed alongside a human being — augmenting judgment, expanding capability, accelerating knowledge — is protected and promoted. AI deployed to eliminate the cost of employing a human being is prohibited. The test is simple: after deployment, are the same workers still employed in the same or better roles, earning the same or more? If yes, the deployment is permitted. If no, the employer must explain why.
02 · The Human-AI Collaboration Mandate
Every covered employer must obtain a HACO-approved Human-AI Collaboration Plan before deploying AI in any covered employment category. The workforce headcount on the date of deployment becomes the permanent minimum — it does not expire. Workers must retain genuine decision authority and override capability, not nominal rubber-stamp review. Algorithmic productivity coercion, constant surveillance as coercion, automated discipline, algorithmic wage manipulation, and psychological manipulation of workers are all prohibited.
03 · Private Equity and Institutional Investor Accountability
PE firms and institutional investors are structurally disqualified from voluntary compliance because their legal obligations, fee structures, and performance benchmarks are built entirely around extraction. Investment agreement clauses that reward headcount reduction, penalize management for maintaining workers, or require AI-driven replacement as a condition of continued investment are void and unenforceable. Every PE firm must audit and submit to HACO a complete accounting of every portfolio company's AI deployment status within 120 days. Joint and several liability attaches to the investor for every violation by a portfolio company it directed, incentivized, or failed to prevent.
04 · Foreign Government Control — The Three-Tier Framework
Tier One: No entity owned, controlled, or materially influenced by any foreign government — regardless of current diplomatic status, alliance membership, or treaty relationship — may deploy AI within the United States or hold operational control over American critical infrastructure AI. No allied-nation exception. No trusted-partner carve-out. Absolute. Tier Two: Foreign privately-owned AI requires a US AI Deployment License — an independent US subsidiary, majority-American board, US-only data storage, 24-hour government override capability, and no foreign parent override. Tier Three: Minority foreign private ownership (5%+) requires registration, annual audits, and annual certification.
05 · Anti-Circumvention and Language Manipulation
The question at every enforcement determination is not what an entity calls what it is doing — it is what the entity is actually doing, measured by observable outcomes. If the workforce shrank after an AI deployment and the employer cannot demonstrate every displaced worker was retained in a comparable role, the deployment is a violation regardless of what the employer calls it. The Euphemism Watch List — maintained publicly by HACO — identifies terms commonly used to disguise prohibited conduct. Law firms, think tanks, and consultants may not be paid to construct arguments whose primary purpose is to narrow the meaning of any term in this policy in a way that would permit conduct the policy is designed to prohibit.
06 · Worker Rights and Transition Protections
No American worker may be displaced by an AI system without: 120 days written notice; 12 months of fully employer-funded retraining in a field of the worker's choosing; 24 months of maintained health insurance for the worker and immediate family; and a transition stipend equal to 100% of annual compensation for the first year. These protections apply to every worker in every U.S. state, territory, and possession — Puerto Rico, Guam, the U.S. Virgin Islands, the Northern Mariana Islands, and American Samoa — on identical terms. The right to collectively refuse AI deployment is affirmed as protected activity under the National Labor Relations Act.
07 · Ecological and Animal Protection
The extractive mandate does not stop at human workers. AI optimization of factory farming that treats animals exclusively as production units without legally binding welfare constraints is prohibited. AI-driven resource extraction exceeding sustainable yield limits is prohibited. AI supply chain optimization that externalizes ecological costs onto natural systems and future generations is prohibited. AI circumvention of environmental law — NEPA, the Endangered Species Act, the Clean Water Act, the Clean Air Act — is prohibited. Civil penalties up to $1,000,000 per violation per day. Mandatory full ecosystem restoration at the violating entity's expense. The protection of workers, animals, and ecosystems are expressions of one principle. This policy addresses all three simultaneously.
08 · The Constitutional Amendment
The Executive Order is the floor. The American Worker Protection and Human-AI Collaboration Act converts it to durable statute. And the constitutional amendment process begins on Day One — establishing the right to human labor, the right to human governance, and the prohibition on AI-driven manipulation in the workplace as permanent constitutional protections. The question of what role machines play in human society should be answered by the American people through their representatives — not by boardrooms, not by algorithms, and not by foreign capital.
Implementation
Executive Order signed. Foreign government bans, anti-circumvention provisions, offshore loophole closures, and emergency powers take effect immediately in all 50 states and all territories.
Portfolio company audit submissions due. Every PE firm and institutional investor must file a complete accounting of every portfolio company's AI deployment status, workforce levels, and investment agreement terms.
Healthcare algorithmic denial prohibition takes effect. No insurance company may deny a medically necessary claim based solely on automated review without a licensed physician signing the denial.
Attorney General constitutional amendment draft due to Congress. Territory-specific HACO implementation guidance published. State grant preference criteria revised.
HACO fully operational. Human-AI Collaboration Plans required for all new AI deployments across all jurisdictions including territories.
Full compliance required for all existing AI deployments. No deployment causing material reduction in human employment may continue without satisfying the Human-AI Collaboration Mandate.
Constitutional Foundation
National Emergencies Act (50 U.S.C. § 1601)
Declares mass automated unemployment a national emergency
IEEPA (50 U.S.C. § 1701)
Regulates foreign government control of AI in U.S. commerce
Defense Production Act (50 U.S.C. § 4501)
Directs human labor allocation in critical sectors
FIRRMA (50 U.S.C. § 4565)
Blocks foreign acquisition of AI control points
Commerce Clause (Article I, § 8)
Regulates AI deployment in interstate commerce
Sherman/Clayton Acts (15 U.S.C. § 1, 12)
Antitrust action against coordinated displacement strategies
FTC Act § 5 (15 U.S.C. § 45)
Unfair competition enforcement against PE automation mandates
NLRA § 7 (29 U.S.C. § 157)
Protects collective refusal of harmful AI deployment
ERISA (29 U.S.C. § 1001)
Fiduciary duty enforcement against extraction-focused investors
Article II (Constitution)
Presidential authority to act on national security threats
Companion Doctrine · June 14, 2026
An AI Security Doctrine Built on Collaboration, Not Containment
Core doctrine: America should not answer foreign advancement by restricting American capability. America should answer it by raising American readiness.
In June 2026, the federal government issued an export-control directive requiring Anthropic to suspend access to its most advanced AI models for foreign nationals — a directive whose scope led to broad access being disabled for all users while the company worked to comply. That action drew the security line in the wrong place.
The restriction targeted 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. The capability at issue — having an AI read code to find software flaws — is knowledge that can be copied, rebuilt, and diffused. Restricting it at one domestic builder slows lawful, visible Americans while leaving every actor who ignores the rules untouched.
Capability denial fails for software knowledge that can be copied, rebuilt, and diffused. If America gates Americans while other nations continue advancing, denial becomes self-harm: the capability still spreads, but Americans fall behind.
Diffusible Capabilities
Code, algorithms, model techniques, prompts, workflows, and knowledge cannot be durably contained. The marginal cost of copying approaches zero. Restriction here mostly penalizes the compliant while actors who ignore the rules proceed unaffected.
Chokepoint-Gated Capabilities
Capabilities requiring scarce physical inputs, specialized equipment, controlled materials, or a very small number of production facilities can sometimes be meaningfully constrained. A serious doctrine distinguishes the two and is honest about it.
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. Restriction loses that race by construction: it slows the defenders who operate in the open while leaving untouched the actors who never asked permission.
During Project Glasswing, Anthropic reported that roughly 50 partners used its restricted model to identify more than 10,000 high- or critical-severity vulnerabilities across important software. Those numbers are not an argument for carelessness. They are an argument for scaling defense faster than offense.
Tier by capability, not by identity.
Most of what makes these tools transformative carries no catastrophic risk and should be widely available. The genuinely dangerous slice is narrower and more specific. Restrict that slice at the capability tier for everyone — with transparent standards and vetting for the highest-risk uses — rather than using nationality or company identity as a substitute for technical judgment.
Universal risk literacy.
Broad access should come with broad understanding. The public should be trained to recognize accident-shaped dangers: data exposure, over-trust in outputs, mishandling sensitive results, false certainty, unsafe automation, and weak review practices. Literacy is a real mitigation for the well-intentioned, but it must be paired with hardened systems.
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, vulnerability remediation, secure-by-design software, resilient infrastructure, and rapid patching converts "the capability exists" from a crisis into a manageable condition.
Narrow, enforceable limits at real chokepoints.
Where a genuine physical bottleneck exists, a limit can be held and may be worth holding. The discipline is to reserve hard restriction for those cases, to review them regularly, and to refuse the comforting illusion that fencing a diffusible capability at a domestic builder makes anyone safer.
No AI capability restriction should be imposed unless the government can satisfy a clear public standard:
This standard does not eliminate government authority. It disciplines it.
The complete policy, executive order, companion legislation, and policy briefs are all available for download. This is not a vision document. It is an operational framework with binding legal standards, enforcement mechanisms, and a constitutional amendment process that begins on Day One.