President Donald Trump has accused Anthropic of endangering troops and jeopardizing national security, but CEO Dario Amodei said his company is patriotic.<\/p>\n
In an interview with CBS News<\/a> soon after Trump ordered the federal government to stop working with Anthropic<\/a>, Amodei pointed out that the AI startup was the first to serve the defense community in a classified setting.<\/p>\n \u201cI believe we have to defend our country from autocratic adversaries like China and like Russia,\u201d he said. \u201cAnd so we\u2019ve been very lean forward. We have a substantial public sector team.\u201d<\/p>\n While Anthropic has provided its AI to the government, the Pentagon demanded unfettered use in all legal scenarios. But the company maintained it has \u201cred lines,\u201d namely its use in domestic mass surveillance and autonomous weapons.<\/p>\n Talks failed to produce an agreement, leading Trump to ban Anthropic from government agencies, while giving the Pentagon a six-month phaseout period. <\/p>\n Defense Secretary Pete Hegseth also called the company a \u201csupply-chain risk,\u201d meaning other contractors working for the Pentagon would not be allowed to use Anthropic\u2019s AI for military work.<\/p>\n Amodei told CBS that Anthropic is onboard with 98%-99% of the military\u2019s use cases. But his concern with mass surveillance is that the latest AI is a game-changer, even within current legal bounds.<\/p>\n \u201cThat actually isn\u2019t illegal. It was just never useful before the era of AI. So there\u2019s this way in which domestic mass surveillance is getting ahead of the law,\u201d he explained. \u201cThe technology\u2019s advancing so fast that it\u2019s out of step with the law.\u201d<\/p>\n As for autonomous weapons, Amodei said AI isn\u2019t reliable enough to take humans completely out of the loop, pointing to the technical problem of \u201cbasic unpredictability\u201d in today\u2019s models.<\/p>\n So far, he is not aware of any real-world examples of a user running up against Anthropic\u2019s red lines but acknowledged that it\u2019s not tenable over the long term for a private company to decide these issues.<\/p>\n Ultimately, Congress must set guardrails on AI\u2019s use, but lawmakers are slow to act, Amodei pointed out. The company is also \u201cnot categorically against fully autonomous weapons,\u201d but believes AI\u2019s reliability isn\u2019t there yet.<\/p>\n In the meantime, Anthropic is still open to working with the government and suggested both sides remain in contact. <\/p>\n \u201cWe are willing to provide our models to all branches of the government, including the Department of War, the intelligence community, the more civilian branches of the government under the terms that we\u2019ve provided under our red lines,\u201d he said.<\/p>\n