The central argument is that frontier AI leaders who believe advanced AI could destroy humanity have a duty not merely to build it more safely than competitors, but to change the conditions driving the race. The author says Anthropic CEO Dario Amodei fails that duty by softening the danger in public. In September interviews, when asked directly whether AI could kill everyone or what the risk was, Amodei redirected toward benefits, government cooperation, or a former employee’s criticism. The author contrasts this with Amodei’s more alarming comments to technical audiences, including a 2023 prediction that dangerous, agentic models might arrive within a few years and be nearly impossible to stop. The mismatch, the author argues, reassures the public at the moment it most needs candor. Other AI executives are criticized as well, with Sam Altman and Demis Hassabis said to communicate even worse.
The piece argues that companies’ actions likewise fall short: they do not plainly warn the public, push for strong international coordination, or use their resources to find lawful ways to slow development. Instead, they emphasize benefits and favor weaker regulation. To demonstrate what genuine responsibility would look like, the author proposes public halt criteria, greater transparency, concrete alignment plans, and pause drills. Companies should also create dedicated public-communications teams, encourage employees to speak openly, prominently warn users, amplify researchers’ concerns, and publicly urge a rapid international moratorium and urgent US-China negotiations. The standard is whether leaders behave as though the stakes are truly existential, even when doing so imposes costs on their businesses.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.