Anthropic, an AI company, has firmly stated that the Pentagon's 'final offer' on AI safeguards is unacceptable. CEO Dario Amodei expressed that the company cannot accept the terms, which defense officials have labeled as their final proposal. The deadline for Anthropic to decide on the Pentagon's offer is fast approaching, and the stakes are high. The Pentagon is threatening to blacklist Anthropic as a supply chain risk or even invoke the Defense Production Act to compel them to provide their model without restrictions. However, Anthropic is not backing down, even with significant gaps remaining in the negotiations. The company remains committed to continuing talks, but if the Pentagon decides to offboard them, Anthropic will work to enable a smooth transition to another provider. This high-stakes feud over the limits on the use of AI model Claude is not unique to Anthropic, as negotiations to bring OpenAI and Google into the classified space are also accelerating. The Pentagon's requirement that AI models be offered for 'all lawful purposes' in classified settings is a point of contention, and Anthropic is standing firm in its position.
