Alignment Researchers Launch Nonprofit Sequent to Tackle Superintelligence Safety

Researchers from the UK AI Security Institute's alignment team and the theory-focused startup Timaeus have launched Sequent, a new nonprofit research organization dedicated to developing alignment methods that can provide high-confidence safety guarantees for superintelligent AI systems.

The founders argue that current empirical alignment programs at frontier AI labs are unlikely to deliver sufficient confidence before artificial superintelligence arrives. Sequent's approach is to pursue a portfolio of differentiated research bets—spanning scalable oversight, learning theory, heuristic arguments, game theory, and personas—rather than relying on the reactive methods most labs currently use. The organization aims to find principled reasons for believing alignment observed during training will generalize to real-world deployment scenarios.

Sequent plans to grow to 40-80 full-time employees within a couple of years, with an initial fundraising target of $100-150M and the possibility of raising substantially more if the parallel research tracks show promise. The group believes pursuing multiple directions simultaneously could yield valuable cross-pollination between approaches.

The organization sees itself as filling a critical gap: building alignment techniques strong enough to withstand recursive self-improvement cycles, while maintaining the independence to raise alarms if frontier labs pursue dangerous paths. "We might need to yell," Sequent's founders write.