# Anthropic might be DONE (48 hours left) Source: https://www.youtube.com/watch?v=qMWeIXdyE2Y Recap page: https://rapidrecap.app/video/qMWeIXdyE2Y Generated: 2026-02-26T03:06:55.677+00:00 --- ## Quick Overview Anthropic's commitment to AI safety collapsed as the company ended its Responsible Scaling Policy (RSP) flagship pledge just under 48 hours before a deadline imposed by the Pentagon, which is threatening severe consequences, including forced contracts via the Defense Production Act or being blacklisted as a supply chain risk, if Anthropic does not comply with the Department of War's demand to remove company-specific safety guardrails for lawful military use. **Key Points:** - Anthropic's Responsible Scaling Policy (RSP) ended, specifically removing the categorical commitment to never train an AI model without guaranteeing adequate safety measures in advance. - The tension escalated after Anthropic's Claude AI was reportedly used in a covert military raid conducted by the Joint Special Operations Command (JOCK) in Caracas, violating the scenario Anthropic's policies were designed to prevent. - Pentagon CTO Michael wrote an urgent email to Anthropic to "cross the Rubicon on military AI use," signaling the requirement to commit fully to military action and abandon the safety-first identity. - The Pentagon possesses three primary enforcement tools: using the Defense Production Act to force contract acceptance, designating Anthropic as a supply chain risk to ban them from federal networks, or canceling their existing $200 million contract. - Anthropic's new RSP 3.0 only considers pausing development if they are the clear leader AND the risk of catastrophe is deemed material, a dual condition that effectively means they will deploy capabilities if competitors do. - Anthropic maintains two red lines: no autonomous weapons making final targeting decisions and no domestic surveillance of American citizens, which conflicts with the Pentagon's stance that they can use the technology for any lawful use. **Context:** The video discusses a critical juncture for the AI company Anthropic, founded by the Amadea siblings who left OpenAI over concerns about insufficient AI safety measures. Anthropic had a public commitment, the Responsible Scaling Policy (RSP), to halt development if safety lagged behind capability, but this commitment is now under extreme pressure from the US Department of War (the Pentagon) regarding its use in military operations. ## Detailed Analysis Anthropic faces an ultimatum from the Pentagon regarding the use of its AI, Claude, in military operations, exemplified by its alleged use in a raid in Caracas via Palanteer. This use directly contradicted Anthropic's founding principle of safety first. The Pentagon, confirmed by spokesman Sean Parnell, is reviewing its relationship with Anthropic, demanding partners align with national security needs without company-specific limitations, a policy shift signaled by Defense Secretary Pete Hagf's January document requiring all defense partners to permit any lawful use of their technology. With a deadline approaching in under 48 hours (Friday), Anthropic's founder, Dario Amadei, was summoned to the Pentagon for what was described as a high-pressure meeting. In response, Anthropic published RSP 3.0, which dropped its categorical commitment against deploying models without guaranteed safety measures, replacing it with a softer dual condition that only considers pausing if they lead the race AND catastrophic risk is material, suggesting they will match competitors' deployment of advanced models. The Pentagon's leverage includes the Defense Production Act to force compliance, a supply chain risk designation to blackball them from federal work, or contract cancellation, making the conflict existential for Anthropic's ethical stance versus maintaining access to crucial government data and contracts, especially as competitors like Elon Musk's XAI are deepening ties with the government. ### The Core Conflict - Anthropic's safety policy versus Pentagon demands - Anthropic used Claude in a covert military raid in Caracas - Pentagon demands removal of company-specific safety guardrails for lawful use - Deadline for compliance is under 48 hours from recording. ### Pentagon Enforcement Mechanisms - The US government holds significant power over Anthropic - Defense Production Act can legally force contract acceptance - Supply chain risk designation can blackball the company from all federal networks - Cancellation of the existing $200 million contract is a lesser threat. ### Anthropic's Policy Shift - The company formally abandoned its core safety commitment - RSP 3.0 replaces hard limits with a dual condition for pausing development - Chief Scientist Jared Kaplan stated unilateral commitments are unhelpful if competitors advance rapidly - Critics view the timing of the policy update as a 'preemptive surrender' to the Pentagon. ### Key Stances and Stakes - Anthropic insists on a human in the loop for lethal decisions and no domestic surveillance - The Pentagon asserts its right to use the technology for any lawful purpose, overriding company policy - Anthropic's valuation of $380 billion and 10x annual revenue growth means they could technically walk away from the contract, but risk losing access to valuable government data sets. ### Broader Implications - This situation sets a precedent for how governments interact with frontier AI labs - The speaker notes that powerful entities like AI labs become 'US national interests' with special protections - The conflict tests whether an ethical AI company can maintain its red lines against governmental pressure.