I also work at an AI company and believe there's plausibly a ≥10% chance that a future out-of-control AI kills everyone (IMO even 1% is unacceptably high). And higher still is the risk that we "only" get permanently disempowered by AIs that don't deeply want the best for us.