Google Steps In Where Anthropic Refused
Google signed a Pentagon AI deal on April 28. Gemini, classified networks, all lawful applications. The catch is that Anthropic was the original supplier and refused to allow domestic mass surveillance and autonomous weapons. The DoD branded Anthropic a supply chain risk for the refusal. A federal judge granted Anthropic an injunction last month. The lawsuit is still active. Google walked through the open door.
This is the third major lab to take an unrestricted Pentagon contract. OpenAI and xAI signed earlier. Anthropic is now the only frontier lab refusing the all-lawful-purposes language. Inside Google, about 950 employees signed an open letter asking the company to adopt Anthropic's stance. The letter did not change the deal.
The twist landed today. Axios reported on April 29 that Trump officials are drafting an executive action that would route around Anthropic's supply chain risk designation and let agencies onboard Anthropic's Mythos model. White House Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent met Dario Amodei earlier in April. The NSA is already using Mythos. Agencies want it. The administration that blacklisted Anthropic in February is now writing the workaround.
What actually happened over six weeks: the government picked a fight with the only AI lab that drew red lines on weapons and surveillance, then quietly realized the model is too good to keep out of the building. For agent builders the takeaway is that procurement gates are being rewritten in real time. Sovereign AI is no longer about national champions versus foreign labs. It is about which labs accept which use cases, and which governments can stomach the labs that say no.
Link: https://techcrunch.com/2026/04/28/google-expands-pentagons-access-to-its-ai-after-anthropic-refusal/
← Back to all articles
This is the third major lab to take an unrestricted Pentagon contract. OpenAI and xAI signed earlier. Anthropic is now the only frontier lab refusing the all-lawful-purposes language. Inside Google, about 950 employees signed an open letter asking the company to adopt Anthropic's stance. The letter did not change the deal.
The twist landed today. Axios reported on April 29 that Trump officials are drafting an executive action that would route around Anthropic's supply chain risk designation and let agencies onboard Anthropic's Mythos model. White House Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent met Dario Amodei earlier in April. The NSA is already using Mythos. Agencies want it. The administration that blacklisted Anthropic in February is now writing the workaround.
What actually happened over six weeks: the government picked a fight with the only AI lab that drew red lines on weapons and surveillance, then quietly realized the model is too good to keep out of the building. For agent builders the takeaway is that procurement gates are being rewritten in real time. Sovereign AI is no longer about national champions versus foreign labs. It is about which labs accept which use cases, and which governments can stomach the labs that say no.
Link: https://techcrunch.com/2026/04/28/google-expands-pentagons-access-to-its-ai-after-anthropic-refusal/
Comments