Focused engagements for teams responsible for safety, governance, and long-horizon strategy at frontier AI organizations.
Facilitated sessions that treat failure as a planning input. We construct plausible catastrophic trajectories, identify the decision points that close them off, and pressure-test whether current controls actually bind.
Independent groups trained to search for the line of play that defeats the current plan. Useful before launches, capability jumps, governance redesigns, and high-stakes external commitments.
Compact, high-signal conversations for safety leads, research directors, and boards. The aim is clarity on stakes, option sets, and where conventional analysis is likely to be overconfident.
Written or working-session input on how a risk is being framed, what is missing from the tree of scenarios, and where evaluation design may be measuring the wrong thing.
We are not a general management consultancy, a communications shop, or a substitute for internal technical safety research. We add strategic and adversarial pressure from outside the building — then leave the implementation to the teams accountable for it.
Grandmaster Labs