Loading…
The Rome Declaration for an Unarmed and Disarming Peace was signed at the Senatorial Palace on Rome's Capitoline Hill during the Global Nobel Laureates Assembly 2026.
More than 30 Nobel laureates signed, alongside AI scientists, religious leaders, philosophers and policymakers.
Its central demand is an international treaty ensuring that no automated system is ever permitted to make the final decision to launch a nuclear weapon.
The declaration warns of a 'convergent peril' arising from delegating moral and life-or-death judgements to AI in nuclear command and control, autonomous weapons and military decision-making.
It insists on 'meaningful human control' as the governing principle, and is structured around six principles and five mandates.
Nuclear deterrence has always depended on decision time. A leader receiving warning of an incoming strike has minutes to judge whether it is real, and the entire architecture of deterrence assumes a human being occupies those minutes. Introducing AI into early warning, target identification and command support compresses that window: machine analysis is faster than human judgement, so the pressure to defer to it grows precisely when the stakes are highest. The declaration's phrase 'convergent peril' names the point where two separately manageable technologies become jointly catastrophic - not because an AI would decide to start a war, but because automation erodes the human pause on which the whole system silently relies. History supplies the argument: false alarms in nuclear early-warning systems have occurred more than once, and were defused by individuals who judged the data implausible and declined to escalate.
Simple Analogy: It is the difference between a smoke alarm that wakes you so you can look, and one wired directly to the sprinklers - the second is faster, and it will flood the house over burnt toast.
Discussed at the UN under the Convention on Certain Conventional Weapons (CCW) through a Group of Governmental Experts. 'Meaningful human control' is the central contested phrase in those talks, and the Rome Declaration adopts it - which is what ties this text to an existing multilateral process rather than leaving it as an isolated appeal.
The NPT (1968) rests on non-proliferation, disarmament and peaceful use; the CTBT (1996) bans test explosions but has never entered into force; the TPNW (2017) prohibits nuclear weapons outright but no nuclear-armed state has joined. India is not a party to the NPT, CTBT or TPNW, maintains a No First Use posture with a credible minimum deterrent, and has a voluntary moratorium on testing.
The strands to cite together: the UNESCO Recommendation on the Ethics of AI (2021), the EU AI Act, the Bletchley Declaration on AI safety (2023), the Global Partnership on Artificial Intelligence, which India has chaired, and India's own approach through the IndiaAI Mission and NITI Aayog's Responsible AI principles.
The Russell-Einstein Manifesto of 1955 and the Pugwash Conferences that followed - awarded the Nobel Peace Prize in 1995 - are the direct precedent for scientists collectively warning against weapons they helped make possible. The Rome Declaration sits in that lineage.
GS Paper 2 > Important international institutions, agreements and their mandate; GS Paper 3 > Science and technology - developments in AI, and internal/external security
Consider the following countries: 1. China 2. France 3. India 4. Israel 5. Pakistan Which among the above are Nuclear Weapons States as recognized by the Treaty on the Non-Proliferation of Nuclear Weapons, commonly known as Nuclear Non-Proliferation Treaty (NPT)?
Answer: 1 and 2 only
With reference to 'Organization for the Prohibition of Chemical Weapons (OPCW)', consider the following statements: 1. It is an organization of European Union in working relation with NATO and WHO. 2. It monitors chemical industry to prevent new weapons from emerging. 3. It provides assistance and protection to States (Parties) against chemical weapons threats. Which of the statements given above is/are correct?
Answer: 2 and 3 only
The principle that a human being must retain genuine, informed decision-making authority over the use of force, rather than merely approving a machine's recommendation.
The systems and procedures through which a state authorises, directs and executes the use of its nuclear weapons.
Weapons that can select and engage targets without human intervention; debated at the UN under the Convention on Certain Conventional Weapons.
The declaration's term for the compounded danger created when AI decision-making is combined with nuclear weapons, exceeding the risk of either alone.
The Treaty on the Prohibition of Nuclear Weapons, 2017, which bans nuclear weapons outright; no nuclear-armed state is a party, and India is not a signatory.