Anthropic Versus Anthropic

Substack · Watch/Read · (gathered 2026-10-29)

• Anthropic exhibits an internal “Anthropic Versus Anthropic” conflict, making costly sacrifices for safety in areas like biosecurity and China policy (often driven by CEO Dario Amodei’s personal priorities) while appearing less rigorous or prepared for other critical risks like AI alignment and loss of control, despite rapidly pursuing recursively self-improving AI (RSI).
• Dario Amodei defends Anthropic’s “race to the top” strategy, citing actions like supporting SB53, strong biological safeguards, and a new Public Benefit Fund (committing 5% of shares). However, critics like Nathan Calvin argue this contradicts earlier statements and that Anthropic’s actions, such as its leaked draft bill, often prioritize preventing government overreach over reducing existential risk.
• Anthropic has granted “employee-like access” to METR (a non-profit) for independent safety evaluations, a significant step for transparency. However, concerns remain about the voluntary nature of such arrangements and an internal “alignment problem” where Anthropic’s policy team often doesn’t reflect the preferences of technical staff or leadership, prioritizing business interests over safety.
• A key warning is that while Anthropic believes RSI is imminent, there is a noted lack of a concrete “plan” for superintelligence alignment, avoiding concentration of power, or envisioning a positive future, beyond hoping future advanced AIs can help solve these issues. The company’s shift towards relying on its own judgment rather than hard external constraints is seen as concerning given its proximity to AGI.