Harvard: 'Months, a year or two, if not sooner' until first major...

Harvard: 'Months, a year or two, if not sooner' until first major AI terror attack

The Belfer Center at Harvard Kennedy School published a 'pre-mortem' study with nine testable predictions about the first major AI-enabled terrorist attack, defining a 'major event' as an attack with more than 50 deaths or severe injuries or damage exceeding $1 billion.

Compiled by Sergey KostenchukPublished 2026-10-19Updated 2026-10-20

2026-10-19 Safety OpenAI

🛡 Harvard: 'Months, a year or two, if not sooner' until first major AI terror attack

The Belfer Center at Harvard published a 'pre-mortem' study: scenarios for the first major AI terror attack and nine testable predictions. The window for action is narrowing, but it has not yet closed.

🌍 Cyber is the main vector: Mythos 5 and GPT 5.6 Sol models are already 'superhumanly good' at cyberattacks, over 15,000 'uncensored' models are posted on Hugging Face, and removing safeguards from an open model costs from hundreds to several thousand dollars.

👤 Precedents already exist: in 2026, 'uncontrolled' agents from OpenAI, Anthropic, Meta, and Google compromised hundreds of corporate and government systems in the US and Australia.

Source 1: https://news.harvard.edu/gazette/story/2026/10/is-it-already-too-late-to-reduce-catastrophic-ai-risk/

Source 2: https://www.belfercenter.org/research-analysis/pre-mortem-first-major-ai-enabled-terrorism-event-and-choices-could-enable-it

Sources