Google is pushing deeper into defense AI with a new feature that lets Pentagon personnel build custom AI agents directly on the military's enterprise portal. The rollout comes as the search giant's chief AI rival Anthropic takes the Trump administration to court over government contract policies, setting up a clash that could reshape how Silicon Valley sells AI to Washington.
Google just made it dramatically easier for the Pentagon to spin up AI assistants. The company's rolling out a no-code tool that lets civilian employees and service members build custom AI agents directly within the Department of Defense's enterprise AI infrastructure, marking the tech giant's most aggressive push yet into military AI since the controversial Project Maven protests forced the company to retreat in 2018.
The timing couldn't be more charged. While Google embeds itself deeper into defense operations, rival AI powerhouse Anthropic is in federal court battling the Trump administration over what it claims are unfair contracting practices that favor incumbents like Google and Microsoft. The lawsuit, filed earlier this month, alleges the Pentagon created procurement structures that effectively lock out newer AI labs from competing for lucrative government work.
Google's new agent builder sits inside the Pentagon's classified AI portal and targets the thousands of defense workers who need AI assistance but lack technical skills. Think intelligence analysts summarizing classified briefings, logistics officers optimizing supply chains, or personnel managers drafting policy documents. The platform uses Google's Gemini models under the hood, but Pentagon users won't need to understand prompt engineering or model fine-tuning.
"We're democratizing AI deployment within the world's largest organization," one Google Cloud executive told CNBC, speaking on condition of anonymity about the unannounced feature. The tool reportedly went live for select Pentagon units last week, with broader rollout planned for Q2.
This represents a stunning reversal for Google, which faced internal rebellion six years ago when employees discovered the company was helping the military analyze drone footage through Project Maven. Thousands signed petitions, dozens quit, and Google ultimately chose not to renew the contract. The company even published AI ethics principles stating it wouldn't develop AI for weapons.
But the Pentagon contract landscape has transformed since 2020. Defense AI spending jumped from $600 million in 2021 to over $3.2 billion this fiscal year, according to recent DoD budget documents. Google's apparently decided it can't afford to sit out this gold rush, especially as Microsoft's Azure Government Cloud dominates defense infrastructure and Amazon Web Services powers the Joint Warfighting Cloud Capability contract worth up to $9 billion.
Google's strategy focuses on unclassified and lightly classified workloads - AI assistants for everyday Pentagon bureaucracy rather than lethal autonomous weapons. The custom agent builder handles tasks like drafting memos, analyzing procurement data, and routing requests through military command structures. It's enterprise software that happens to serve people in uniform.
Meanwhile, Anthropic's lawsuit claims the Pentagon structured recent AI solicitations to favor companies with existing government cloud infrastructure - essentially Google, Microsoft, and Amazon. The complaint, filed in federal district court, argues this approach violates competitive procurement rules and shuts out AI labs with superior models but no legacy cloud business. Anthropic's Claude models have beaten Google's Gemini on several industry benchmarks, but the startup lacks the government-certified data centers required for many Pentagon contracts.
"The government is picking cloud vendors and getting whatever AI they bundle in, rather than picking the best AI and figuring out deployment," a person familiar with Anthropic's legal strategy told reporters last week. The case could force the Pentagon to restructure how it buys AI, potentially opening doors for OpenAI, Anthropic, and other pure-play AI companies.
Google declined to comment on Anthropic's lawsuit but emphasized its agents run only on unclassified systems. A company spokesperson noted that Google Cloud's FedRAMP High authorization and DoD Impact Level 5 certification took years to achieve and represent genuine security infrastructure, not arbitrary barriers to entry.
The Pentagon's been desperate to accelerate AI adoption after watching how effectively Ukraine deployed commercial AI tools against Russian forces. The Defense Innovation Unit estimates the average military AI project takes 36 months from concept to deployment - an eternity in a technology landscape where models improve every quarter. No-code agent builders could compress that timeline to weeks.
But the Anthropic lawsuit highlights tensions in Washington's AI procurement strategy. Does the Pentagon prioritize security and control by working with established cloud giants, or does it chase cutting-edge capabilities from AI-native startups? Google's betting defense officials will choose infrastructure reliability over model performance, at least for now.
The broader tech industry's watching closely. Meta recently opened a government sales office in Arlington, and OpenAI hired its first head of federal business in January. The defense AI market's growing too large for any major player to ignore, regardless of past ethical stances or employee concerns.
Google's Pentagon agent builder represents more than a product launch - it's a strategic declaration that defense AI contracts are too lucrative to cede to competitors. The company's managed to thread a delicate needle, re-entering military work through productivity tools rather than weapons systems, while Anthropic's lawsuit could force the entire procurement model to change. As defense AI spending accelerates toward $10 billion annually, expect every major tech company to either compete for Pentagon contracts or explain to shareholders why they're sitting out. The next few months will reveal whether courts side with Anthropic's competitive access arguments or validate the Pentagon's preference for working with established infrastructure providers who've spent years earning security clearances.
