A recent letter signed by 1,367 researchers and engineers at frontier AI labs – mainly OpenAI, Anthropic and Google Deepmind – points to a dangerous moment