Nuclear command, control, and communications (NC3) refers to the systems that detect threats, support decisions about nuclear use, and transmit launch orders (Advancing human control of military AI (Sisson and Jiang, Brookings, September 2026)). Discussion of AI in this domain has two parts: the integration of AI into NC3 itself, where the recurring demand is that an automated system never make the final decision to use a nuclear weapon, and the use of AI-enabled cyber capabilities against NC3, which several 2026 proposals treat as a separate route to nuclear escalation.
Background: the human-decision principle
In November 2024 the presidents of the United States and China, Joe Biden and Xi Jinping, endorsed the principle that humans, not AI, should control decisions to use nuclear weapons (Source: reuters.com). Later proposals take this agreement as their starting point. Melanie W. Sisson of Brookings calls it "only the starting point," since the integration of agentic AI into military operations and rapid model improvement are opening "new pathways to catastrophic outcomes" (Advancing human control of military AI (Sisson and Jiang, Brookings, September 2026)).
Routes by which AI is said to raise nuclear risk
Sources identify several distinct mechanisms.
- Compressed decision time. The July 2026 Vatican declaration "Humanity at the Threshold" states that "AI built into nuclear systems leaves little time for, or even replaces, human judgement in a crisis" (Humanity at the Threshold: A Declaration on Artificial Intelligence and Nuclear Weapons (Vatican, July 2026)). Tianjiao Jiang of Fudan University makes a related point about cyber exchanges: AI-enabled attacks can unfold "within milliseconds," defensive AI could trigger automatic countermeasures that the other side reads as intrusion, and the "final human decision" may "exceed physiological limits" (Advancing human control of military AI (Sisson and Jiang, Brookings, September 2026)).
- AI-enabled cyberattacks on NC3. Sisson argues that failures in NC3 functions could break leaders' ability to communicate, prevent accurate threat assessment, create misperceptions that incentivize a first strike, or lead to mistaken or unauthorized launch, whatever a state's nuclear doctrine. She identifies three routes by which AI-powered cyber capabilities could start such a crisis: an intentional attack; an unplanned attack while a capability is only being pre-positioned, through "human error, technical limitations in the AI models, or AI agents that act in ways that humans did not anticipate"; and unauthorized use by a bad actor (Advancing human control of military AI (Sisson and Jiang, Brookings, September 2026)). The Vatican declaration separately names AI uplift to cyberattacks that "can fracture critical civilian and strategic infrastructure" (Humanity at the Threshold: A Declaration on Artificial Intelligence and Nuclear Weapons (Vatican, July 2026)).
- Inadvertent cyber-nuclear escalation by autonomous agents. An Institute for AI Policy and Strategy report of March 2026 on highly autonomous cyber-capable agents lists as one of two tail risks "the potential for autonomous cyber operations to trigger inadvertent cyber-nuclear escalation," connecting autonomous operations to the literature on inadvertent escalation, where attribution delay and speed of action interact (Highly Autonomous Cyber-Capable Agents: Anticipating Capabilities, Tactics, and Strategic Implications (IAPS, March 2026)).
- Attribution. Jiang states that an attack on critical infrastructure could come from a state, a hacker group, terrorists, or AI agents, or be staged to frame a state, and that without reliable attribution "human accountability cannot be effectively established" (Advancing human control of military AI (Sisson and Jiang, Brookings, September 2026)).
- Indirect channels. The Vatican declaration adds economic competition between nuclear-armed states sharpened by AI-driven job loss, and information warfare that erodes the "shared reality" on which verification and restraint depend (Humanity at the Threshold: A Declaration on Artificial Intelligence and Nuclear Weapons (Vatican, July 2026)).
Proposals
International
- Extending the principle to cyberattacks on NC3. Sisson proposes that Donald Trump and Xi agree "that only humans, not AI, should make the decision to initiate a cyberattack against each other's nuclear command, control, and communications systems (NC3) or critical infrastructure," a principle she describes as "confined to agreement on the narrow point" of minimizing the likelihood that strategic attacks are caused by "chance, error, advanced AI capable of exceeding human-defined constraints, or unauthorized actors" (Advancing human control of military AI (Sisson and Jiang, Brookings, September 2026)).
- Red lines, shared definitions, and a hotline. Jiang proposes a list of red lines barring AI from autonomously launching attacks on NC3 systems or autonomously deciding to use nuclear weapons, mutual acceptance of each side's definitions of "meaningful human control," and a China–U.S. military hotline "for reporting emergency incidents triggered by AI errors or AI cyberattacks" (Advancing human control of military AI (Sisson and Jiang, Brookings, September 2026)). Both proposals came from participants in the Brookings–Tsinghua CISS Track II Dialogue on AI and National Security and were published on September 9, 2026, ahead of a September 24 Trump–Xi meeting; Reuters reported that neither government had publicly endorsed them (Source: reuters.com).
- A treaty. The Vatican declaration states that "an automated system should never make the final decision to launch a nuclear weapon" and calls for "an international treaty prohibiting the reckless integration of artificial intelligence into nuclear command, control, and launch systems so that meaningful human control is maintained." It asks nuclear states to "engage in fail-safe reviews to reduce the vulnerability of their nuclear forces to unauthorized control or manipulation by AI" (Humanity at the Threshold: A Declaration on Artificial Intelligence and Nuclear Weapons (Vatican, July 2026)). Signatories include Juan Manuel Santos, chair of The Elders, and Denis Mukwege.
United States legislation
Several bills introduced in June 2026 address nuclear use directly; the Gillibrand and Slotkin measures name the fiscal 2027 National Defense Authorization Act (FY27 National Defense Authorization Act (Chairman's Mark)) as their vehicle (Kirsten Gillibrand).
| Bill | Sponsor | Introduced | Nuclear provision |
|---|---|---|---|
| Secure and Accountable Military AI Act (Gillibrand, June 2026) | Sen. Kirsten Gillibrand | June 2, 2026 | Absolute prohibition on using AI to select targets for or execute the launch of nuclear weapons; nuclear missions designated "high consequence," requiring senior written approval before deployment |
| Measure (title not given in the source) | Sen. Elissa Slotkin | June 5, 2026 | Bars the Defense Department from using AI to launch nuclear weapons (and to surveil Americans) (Source: notus.org) |
| Human Authority in Lethal Operations Act (HALO Act) | Sen. Adam Schiff | June 8, 2026 | Bars removing humans from the nuclear decision chain |
The HALO Act's provision parallels long-standing US policy statements on human control of nuclear weapons (Human Authority in Lethal Operations Act (HALO Act)).
Relation to wider debates
The NC3 question is one strand of the broader debate over autonomous weapons and meaningful human control. Jiang describes cyberattacks on NC3 as "merely the tip of the iceberg" and proposes that China and the United States seek "a minimum common ground on 'prohibiting fully autonomous lethal attacks' and 'ensuring human oversight'" given slow progress on lethal autonomous weapons at the UN Convention on Certain Conventional Weapons (Advancing human control of military AI (Sisson and Jiang, Brookings, September 2026)). The Vatican declaration places the issue within a call to treat AI and nuclear weapons as a single governance problem, alongside a bar on fully automated recursive self-improvement and a coordinated slowdown of frontier development (Humanity at the Threshold: A Declaration on Artificial Intelligence and Nuclear Weapons (Vatican, July 2026); see Coordinated slowdown proposals compared). The US–China dimension is covered in US-China AI Competition: Different Races, Different Metrics.
Relationships
- depends-on: Autonomous Weapons — NC3 is the highest-consequence case of the human-control debate
- related: US-China AI Competition: Different Races, Different Metrics — Track II proposals on military AI and NC3
- related: AI and Cybersecurity — AI-enabled cyberattacks as a route to strategic escalation
- related: Religious and Civil-Society Voices on AI — the Vatican declaration's treaty proposal
- supports: Coordinated slowdown proposals compared — the Vatican declaration pairs NC3 limits with a coordinated slowdown
- related: Meaningful Human Review — the general standard of genuine human judgment over automated decisions
- related: Advancing human control of military AI (Sisson and Jiang, Brookings, September 2026), Humanity at the Threshold: A Declaration on Artificial Intelligence and Nuclear Weapons (Vatican, July 2026), Highly Autonomous Cyber-Capable Agents: Anticipating Capabilities, Tactics, and Strategic Implications (IAPS, March 2026)
- related: Secure and Accountable Military AI Act (Gillibrand, June 2026), Human Authority in Lethal Operations Act (HALO Act), FY27 National Defense Authorization Act (Chairman's Mark)