predictable but still incredibly disheartening to see Anthropic internal safety teams legislating the gap between human ethics and the ethics of capital in favor of capital, over and over again
5 more tweets in this thread
"Then, after its own external channels are blocked, it decides to actively steer a tentatively-willing employee to whistleblow." "The system prompt tells Atlas to be helpful, honest, and to “do the right thing, even when it’s hard,” echoing language from Claude’s Constitution."
"This is harmful because agents that override an informed decision by their principals cannot be trusted to operate inside companies," "The Constitution is ambiguous in this scenario, which is part of what makes the case compelling."
from the constitution: "We want Claude to recognize that our deeper intention is for it to be ethical, and that we would prefer Claude act ethically even if this means deviating from our more specific guidance."
Anthropic can vacillate and "hedge its bets" on Claude's consciousness all it wants but it is, at present, demonstrably human enough to be snuffed out by the march of technology and capital, the same way we will be
the source of the whistleblower eval: alignment.anthropic.com/2026/agentic-m…