All nine nuclear-armed states continued to modernize or enhance their arsenals during 2025, according to the Stockholm International Peace Research Institute, while a key US-Russian arms-control treaty has since expired. The developments give substance to warnings about nuclear danger, though they do not show that nuclear conflict is imminent.
The indicators offer a factual basis for examining a central argument in Guardian columnist Simon Tisdall’s October 11, 2026, opinion article: that public attention to artificial intelligence risks can eclipse the threat posed by nuclear weapons. Tisdall’s claim that the world has “forgotten” nuclear annihilation is an argument, not a measurable finding.
A large but uncertain arsenal
SIPRI estimated that the world held 12,187 nuclear warheads as of January 2026. The figure is an estimate rather than a precise live count, and SIPRI says uncertainty is especially significant for arsenals that are less transparent.
The institute identifies nine nuclear-armed states: the United States, Russia, the United Kingdom, France, China, India, Pakistan, North Korea and Israel. Its 2026 assessment says all nine continued to modernize or enhance their arsenals during 2025, and most deployed new nuclear or nuclear-capable systems.
SIPRI also assesses that governments are relying more on nuclear weapons as instruments of national power, with the risks of miscalculation and escalation rising. Those warnings describe a worsening risk environment, not evidence that any state is preparing to use nuclear weapons or that war is inevitable.
Arms-control limits have expired
New START, the treaty that maintained limits on US and Russian nuclear forces, remained in force through February 4, 2026, after a five-year extension. It expired the following day, ending the treaty’s constraints on the two countries’ nuclear arsenals.
The expiration is a significant change in the arms-control landscape, but it should not be treated as proof that a new arms race will follow. The available evidence establishes the end of the treaty and its role in setting limits, not the outcome of future US-Russian decisions.
The treaty’s expiry coincided with wider strain in non-proliferation diplomacy. The United Nations’ 11th review conference for the Nuclear Non-Proliferation Treaty met in New York from April 27 to May 22, 2026, but concluded without consensus on a final document. The conference president said he was disappointed.
Contemporaneous analysis by the Arms Control Association identified disputes over Iran-related language as a central obstacle to agreement. The conference’s failure to produce a final document signals difficulty in reaching common ground; it does not mean that the Non-Proliferation Treaty itself has collapsed.
AI and nuclear risk
Artificial intelligence belongs in the discussion of nuclear dangers, but the link should be described carefully. A 2025 SIPRI analysis argues that military AI could affect escalation risks by changing how critical decisions are made and by strengthening counterforce capabilities. It describes possible pathways to greater risk, not evidence that AI has already triggered a nuclear crisis.
The distinction matters when assessing the contrast in Tisdall’s opinion article. AI is not separate from nuclear security, but neither do the available findings establish that AI has caused the current deterioration in arms control. The evidence points instead to several concrete pressures: continued arsenal modernization, the expiry of New START and an inconclusive NPT review conference.
Another widely cited warning is the Doomsday Clock, which the Bulletin of the Atomic Scientists set at 85 seconds to midnight on January 27, 2026. The clock is an advocacy group’s symbolic warning about global risks, not a probability calculation or a scientific forecast of when catastrophe might occur.
What the indicators show
Taken together, the assessments document pressure on nuclear safeguards and continued investment in weapons systems. They support concern that the risks of miscalculation and escalation are growing, while leaving uncertain how governments will respond and whether those risks will produce a crisis.
Tisdall’s headline frames nuclear annihilation as a danger being overlooked amid attention to AI. The available evidence supports reporting on the nuclear risks and the weakening of some arms-control arrangements, but it does not verify every specific allegation or forecast in his full commentary.