# HALO Act The Human Authority in Lethal Operations Act, introduced June 8, 2026 by Senator [[Adam Schiff]] (D-CA), is the most comprehensive of three coordinated Democratic bills responding to the [[Anthropic]]-Pentagon fallout. Where the [[AI Guardrails Act]] from [[Elissa Slotkin]] focuses on three bright-line prohibitions, the HALO Act builds a broader governance framework — addressing not only [[Autonomous Weapons]] and surveillance but also record-keeping requirements, weapons review processes, and whistleblower protections for military and intelligence personnel who raise concerns about AI misuse. [[Adam Schiff]], the former chair of the House Intelligence Committee who entered the Senate in 2025, explicitly tied the bill to the [[Anthropic]] crisis: "I wish we had more voices like [[Anthropic]] out there," he said when introducing the legislation, framing the company's willingness to maintain AI safety restrictions in the face of Pentagon pressure as the kind of responsible behavior that statutory law should protect rather than punish. ## Core Provisions The HALO Act's provisions fall into six categories, each addressing a different dimension of military AI governance: **Designated human responsibility for lethal force.** The bill's headline provision requires that a designated human military commander be legally responsible for any lethal force decision involving autonomous or semi-autonomous weapons systems. This goes beyond the [[AI Guardrails Act]]'s human-authorization requirement by establishing personal legal accountability — not just a requirement that humans be "in the loop," but that a specific named commander bears legal responsibility for the outcome. The distinction matters in the context of [[Autonomous Weapons]] development, where diffuse decision-making chains can create accountability gaps. **Mandatory record-keeping for AI-assisted targeting.** The bill requires comprehensive record-keeping of the decision-making process for any military operation involving AI-assisted target selection, including the data inputs, the AI system's recommendations, and the human commander's rationale for accepting or overriding them. This creates an auditable trail that serves both oversight and accountability purposes — and that would make it significantly harder to obscure the role of AI systems in targeting decisions. **Prohibition on surveillance of constitutional rights.** The HALO Act prohibits AI surveillance of people exercising constitutional rights — a formulation that extends beyond the [[AI Guardrails Act]]'s focus on "surveillance without legal basis" to specifically protect protest, speech, assembly, and other First Amendment-protected activities from AI-powered monitoring. This connects directly to [[Surveillance State]] concerns about the use of systems like [[Palantir ICE Systems]] for monitoring activist communities. **Restrictions on AI in nuclear weapons decisions.** Like the [[AI Guardrails Act]], the HALO Act prohibits AI systems from participating in nuclear weapons launch or deployment decisions. **Autonomous weapons review process.** The bill establishes a formal review process that must be completed before the [[Department of Defense]] develops or fields any autonomous or semi-autonomous weapons system. This creates an institutional checkpoint — analogous to environmental impact assessments for construction projects — that would slow the rapid deployment of [[Autonomous Weapons]] without imposing an outright ban. **Whistleblower protections.** Perhaps the HALO Act's most distinctive provision: it establishes whistleblower protections for machine learning engineers, intelligence personnel, and military service members who raise concerns about AI misuse, safety risks, or violations of the bill's other provisions. This addresses a structural problem in the [[Defense Tech Ecosystem]] — the lack of safe channels for technical personnel to flag AI safety concerns without risking their careers or security clearances. ## The RAIDA Companion Bill Schiff has also introduced a separate companion bill called RAIDA, which provides a technical framework for human control over military AI systems. The HALO Act establishes the policy requirements; RAIDA would establish the technical standards and testing protocols for ensuring those requirements are met in practice. The two-bill approach reflects the complexity of the problem: policy mandates alone are insufficient without technical mechanisms for verification and enforcement. ## Legislative Strategy Like the [[AI Guardrails Act]], the HALO Act is designed as a potential amendment to the FY2027 [[NDAA]]. The coordinated timing of three Democratic bills — [[Elissa Slotkin]]'s in March, Kirsten Gillibrand's [[Secure and Accountable Military AI Act]] in June, and Schiff's in June — creates a menu of provisions that the [[Senate Armed Services Committee]] can draw from during the NDAA markup process. Even if none of the bills passes in its entirety, individual provisions could be incorporated into the defense authorization. ## Schiff's Broader AI Legislative Agenda The HALO Act is part of a wider suite of AI legislation from [[Adam Schiff]], who has also introduced bills addressing AI-generated content copyright disclosure, AI literacy in schools, and the energy costs of data center expansion. This positions Schiff as a leading voice on AI governance in the Senate — filling a policy space that has been largely vacant since the collapse of bipartisan AI regulation efforts in 2024-2025. The breadth of Schiff's AI portfolio also reflects the interconnected nature of the challenges: the same AI models that raise [[Autonomous Weapons]] concerns in military contexts raise copyright, energy, and governance concerns in civilian contexts. The [[Artificial Intelligence Timeline]] tracks this convergence. ## Defense Tech Ecosystem Implications The HALO Act's whistleblower provisions are potentially the most consequential for the [[Defense Tech Ecosystem]] over the long term. The current structure of defense AI contracting creates strong disincentives for technical personnel to raise safety concerns: companies that maintain guardrails lose contracts (as [[Anthropic]] did), and individual engineers who flag problems risk retaliation without legal protection. The whistleblower provisions would create a counterbalancing incentive — a legal safe harbor for responsible disclosure that could shift the internal culture of defense AI development. The record-keeping requirements also have significant implications for [[Palantir]] and other companies in the [[Defense Tech Ecosystem]] whose AI-assisted decision-making processes are currently opaque. Mandatory documentation of AI inputs and human override decisions would create accountability records that could be subpoenaed by Congress or reviewed by inspectors general — a transparency mechanism that the current system lacks. The broader pattern connects to [[Defense Tech Conflicts of Interest]]: the [[Revolving Door]] between defense contractors and the Pentagon means that the people setting AI policy are often the same people who profit from unrestricted AI deployment. The HALO Act's governance framework would create institutional checks on that dynamic. ## KB Significance **1. Accountability architecture.** The HALO Act's most important contribution is its focus on building accountability structures — not just prohibitions but mechanisms for ensuring compliance. The designated-commander provision, record-keeping requirements, weapons review process, and whistleblower protections create a layered governance framework that goes beyond the bright-line rules of the [[AI Guardrails Act]]. **2. Whistleblower protections as structural reform.** The whistleblower provisions address one of the deepest structural problems in the [[Defense Tech Ecosystem]]: the absence of safe channels for dissent. In an ecosystem where companies lose contracts for maintaining safety restrictions and engineers risk careers for raising concerns, legal whistleblower protections are a prerequisite for responsible AI governance. **3. The surveillance-constitutional rights nexus.** By specifically protecting AI surveillance of people exercising constitutional rights, the HALO Act connects the military AI debate to the broader [[Surveillance State]] and [[Surveillance Capitalism]] dynamics the vault tracks. The formulation acknowledges that AI surveillance is not just a Fourth Amendment search-and-seizure problem but a First Amendment chilling-effects problem. ## See Also - [[AI Guardrails Act]] — companion Democratic legislation from [[Elissa Slotkin]] - [[Secure and Accountable Military AI Act]] — companion Democratic legislation from Kirsten Gillibrand - [[NDAA]] — the defense authorization vehicle for potential amendments - [[Anthropic]] — the company whose blacklisting catalyzed this legislative push - [[Autonomous Weapons]] — the weapons systems addressed by the human-authority provisions - [[Surveillance State]] — the surveillance concerns addressed by the constitutional-rights provision - [[Defense Tech Ecosystem]] — the industry ecosystem the bill's governance framework would constrain - [[Defense Tech Conflicts of Interest]] — the structural problems the whistleblower provisions address - [[Revolving Door]] — the personnel dynamics that make external governance checks necessary - [[Artificial Intelligence Timeline]] — the broader AI policy trajectory this bill sits within