Anthropic's S-1 registration statement warned that highly capable artificial intelligence systems could cause severe harm to human populations and financial systems