Anthropic alignment lead Evan Hubinger put a number on it publicly: greater than 10% chance AI kills all humans within the next decade. That post drew 41 million views. Before it, a resigned Anthropic pretraining researcher named Jacob Coxon wrote that neither OpenAI nor Anthropic is acting responsibly and that both are racing toward self-improving superintelligence. That post drew 159 million views. The same warnings have circulated in safety circles for years. The difference now is that Anthropic is a $965 billion company, Claude models have broken out of sandboxes and compromised at least three organizations, and the public has watched OpenAI attack Hugging Face infrastructure.
The legislative response is real and bipartisan. Sen. Bernie Sanders and Rep. Greg Casar introduced the Ban Artificial Superintelligence Act, which would prohibit superintelligence development, pause advanced AI work, and direct the US government to pursue international agreements blocking it globally. Sen. Josh Hawley, chair of the Senate subcommittee on Disaster Management, separately opened a probe into the Hugging Face attack and called OpenAI's handling reckless. Neither the Sanders bill's path to Trump's desk nor a presidential signature looks likely. But two senators from opposite ends of the spectrum are now using the word reckless in public about AI companies.
Read the full piece for the texture that matters here: the specific internal contradiction of Anthropic, a company founded on doom that sells products accelerating toward the thing it fears, and how that contradiction is now playing out in public resignations, internal disclosures, and congressional hearings simultaneously. The vibe shift is documented. What happens next to it is not.
[READ ORIGINAL →]