Question
Will a nuclear weapon be used (intentionally in an attack/war, or accidentally via unauthorized launch or detonation) anywhere in the world between now and December 31, 2027?
The probability of a qualifying nuclear detonation over the next 15 months is estimated at 2%. This assessment balances a deteriorating global security environment against an overwhelmingly strong historical precedent. No nuclear weapon has been detonated in conflict or by accident since 1945, an 81-year record encompassing roughly 970 months of non-use. Crucially, the explicit exclusion of declared state weapons tests removes the single most likely route to a nuclear explosion in this window. Any qualifying event therefore requires either a deliberate decision that breaks a profound global taboo, or an unprecedented failure of redundant safety systems.
Nevertheless, systemic risks are genuinely elevated compared to the post-Cold War norm. The expiration of New START in February 2026 left US-Russia strategic arms control unconstrained for the first time in decades 2 sources. Global stockpiles stand at approximately 12,187 warheads, with roughly 2,100–2,200 on high operational alert sipri.org. Rising risks of miscalculation are compounded by the loss of crisis-management channels, growing strategic ambiguity, reported shifts in nuclear employment guidance emphasizing low-yield weapons thebulletin.org, and the gradual integration of AI into early-warning and command-and-control systems nti.org. The January 2026 setting of the Doomsday Clock to 85 seconds to midnight reflects this historically degraded structural environment thebulletin.org.
The most significant pathway for intentional use remains escalation involving Russia, Ukraine, and NATO. While peace talks are at a dead end and reporting in August 2026 indicates some Russian officials have discussed tactical nuclear use as a "last resort" 3 sources, actual employment is highly unlikely. Moscow is currently pursuing conventional attritional strategies rather than nuclear escalation understandingwar.org, and its repeated threats are best understood as reflexive-control signaling fdd.org. Furthermore, the military utility of a battlefield detonation is low, and Russia faces severe constraints from key partners—particularly China and India—who have explicitly warned against breaking the nuclear taboo 2 sources. Absent an imminent, existential battlefield collapse, Moscow is unlikely to cross this threshold.
Other regional flashpoints contribute marginally to the overall risk but face similar deterrents. In South Asia, despite the severe May 2025 military confrontation and subsequent naval incidents highlighting a lack of crisis hotlines 2 sources, the overarching US-brokered ceasefire continues to hold cfr.org. In the Middle East, ongoing instability is constrained by the fact that Iran's program has been heavily degraded and it currently lacks a deliverable nuclear weapon 2 sources, while an Israeli or US nuclear strike remains highly implausible absent a true existential threat manaramagazine.org. In East Asia, North Korea continues to expand its arsenal and delivery systems 2 sources, but Pyongyang's overriding priority of regime survival heavily deters a first use that would invite catastrophic retaliation.
Finally, the risk of an accidental or unauthorized yield-producing detonation is extraordinarily low. While historical near-misses and early-warning ambiguities demonstrate that false alarms occur, every incident in the nuclear age has been caught before launch sipri.org. Modern nuclear weapons are highly engineered to be one-point safe, meaning conventional accidents do not produce a nuclear yield osti.gov. Although cyber vulnerabilities and AI integration introduce novel failure modes, an accident bypassing all safety mechanisms to produce a fission or fusion yield over a 15-month horizon remains a remote tail risk. Barring a sudden spiral in an active conflict zone, the structural disincentives against nuclear use remain intact.
Ask a followup
Sign in to run · $20 free credit, no card · every claim cited