Issues

Nuclear Risk

12,000+ warheads, thousands still on high alert.

Nuclear weapons are the one existential risk that has already been used in war, and many of the arms control agreements that limited them have lapsed.

We host speakers on deterrence and arms control, and connect students with people working in nonproliferation.

Biological Risk

• Biological risk is the chance that a pathogen, released by accident or on purpose, causes a pandemic severe enough to threaten civilization, and the work of preventing that through detection, preparedness, and oversight of dangerous research.
• States have cheated before, at enormous scale. After signing the 1972 Biological Weapons Convention, the Soviet Union secretly ran the largest bioweapons program in history through Biopreparat, a front organization (Leitenberg & Zilinskas, 2012). At its peak it employed roughly 65,000 people across some 40 to 50 facilities, and a single production building at Stepnogorsk was built to turn out 300 metric tons of weaponized anthrax in a ten-month cycle (Frontiers in Political Science, 2025). A lethal inhaled dose is on the order of ten nanograms (PLOS Pathogens), so that single cycle is, in theory, millions of lethal doses for every person alive — a figure that assumes flawless delivery to each individual and so far overstates any achievable attack, but conveys the scale of what was built.
• The barriers are falling now. Frontier AI developers have begun treating their own models as capable enough in biology to require heightened safeguards, and independent evaluators track these declarations across model reports (SecureBio). This is no longer hypothetical: in 2026 Anthropic reported detecting and blocking attempts to use its models in support of biological weapons work (PBS NewsHour, 2026).
• Non-state groups have already tried. Aum Shinrikyo, the cult behind the 1995 Tokyo subway sarin attack, also pursued biological weapons. It failed, but groups willing to seek casualties on a catastrophic scale keep appearing: of roughly 270 documented cases involving biological agents over the last century, only a handful sought mass casualties, Aum and the 1970s group R.I.S.E. among the clearest (Carus, National Defense University). Advances in AI and synthetic biology are making it increasingly likely that these groups will be able to achieve their goals (Giordano & DiEuliis, mSphere, 2025; Wittmann et al., Science, 2025).

AI Risk

AI capabilities advancing faster than our ability to control them.

Concerns include misuse and the difficulty of controlling systems we don't fully understand.

AI can also make the other two risks worse, for example by making it easier to design a pathogen.

We host people working on AI safety and policy, and help students find ways into the field.