Anthropic Open Letter: The Hypocritical Sam Altman, Master of PUA
Anthropic CEO Dario Amodei accuses OpenAI and Sam Altman of hypocrisy and "safety theater" in a leaked internal memo. The email follows the Pentagon’s termination of its contract with Anthropic over safety terms, while OpenAI secured a deal with the Defense Department hours later.
Amodei claims OpenAI’s model has "no legal use restrictions" and relies on a superficial "safety layer"—such as refusal mechanisms or third-party classifiers—which he deems 80% performative and only 20% effective. He argues these measures fail to prevent misuse in mass surveillance or autonomous weapons, as models lack contextual awareness and are vulnerable to jailbreaks.
He reveals the Pentagon rejected similar safety clauses proposed by Anthropic, particularly one targeting "analysis of bulk acquired data." Amodei alleges the Pentagon and OpenAI coordinated narratives to downplay risks, claiming existing laws already prohibit misuse—a stance he disputes.
Amodei accuses Altman of gaslighting, political maneuvering, and undermining Anthropic’s stance while appealing to OpenAI employees with performative safety pledges. He links OpenAI’s favorable treatment to its political donations and alignment with the Trump administration’s agenda, contrasting it with Anthropic’s refusal to compromise on ethical red lines.
marsbit03/05 05:50
