The 'doomsday warning' has gone viral? Anthropic's chatty risk marketing with real data, offering a great gift package for geographical supply interruption According to Beating monitoring, the Financial Times has conducted a statistical analysis of the wording used by Anthropic and OpenAI on social media, official statements, and articles throughout the year, which has proven Anthropic's "chatterbox". Data shows that in Anthropic's public statements, an average of 5 out of every 1000 words discuss "risk, regulation, or restriction," including 336 mentions of "risk," 121 mentions of "safety," and 128 mentions of "vulnerability. In contrast, the ratio between OpenAI and Sam Altman is only 0.6 words per thousand words, almost 8 times lower. This contrast has led to Anthropic's safety public relations being widely ridiculed by the outside world as' Oppenheimer style marketing '. As early as this April's podcast, Sam Altman openly complained about this routine: "This is obviously an incredible marketing method, shouting around:" We made a bomb, and now we are going to hit you! "This kind of risk marketing, which pretends to be" safe and kind-hearted in the industry ", has finally become a catalyst for self restraint. Amazon, the major shareholder, betrayed the White House for its own interests and reported the bypass vulnerability of Fable 5, which directly led to the White House implementing a 90 minute ultimatum and a forced offline ban. Ironically, it was precisely because Anthropic had repeatedly compared the risks of AI to an "atomic bomb" in front of politicians, and even boasted about Mythos' "cyber weapon level" destructive power, that the White House naturally killed him in this incident and banned him on the grounds of national security. Dario Amodei's painstaking venture marketing ultimately provided the White House with the most convenient leverage to hit him. [Original link]