Bottom Line: Big Tech’s “Divorce” From Defense Is Over
On April 28, 2026, Google signed an agreement allowing the Pentagon to use its AI models for classified work under “any lawful government purpose.”
The significance: the agreement removed previous clauses restricting military applications. Google AI is no longer limited to logistics and data analysis — it can now extend to mission planning and operational use.
Over 600 Google employeesjointly protested, urging CEO Sundar Pichai to reject the deal, but failed to change the company’s decision. This marks a historic turning point since Google exited Project Maven in 2018 due to employee protests.
Event Breakdown
Key Change: From “Limited Participation” to “Full Openness”
| Phase | Time | Restrictions | Use Cases |
|---|---|---|---|
| Exit period | 2018 | Refused participation | Google exits Maven due to employee protests |
| Limited period | 2019-2025 | Military use restricted | Only logistics, non-combat analysis |
| Full openness | Apr 2026 | No restrictions | Mission planning, operations, “any lawful purpose” |
The phrase “any lawful government purpose” is the core change — from explicitly excluding military use to not excluding any use.
Three-Company Landscape
Google’s entry makes defense AI a three-horse race among Big Tech:
| Company | Defense AI Status | Key Contracts |
|---|---|---|
| OpenAI | Signed classified contracts | Multiple DoD AI cooperation agreements |
| xAI | Signed classified contracts | Grok model defense applications |
| New entrant | Gemini model classified work support |
Why It Matters
1. Technology Spillover Effects
Defense contracts impose requirements far beyond commercial use cases:
- Extreme reliability: Errors cost lives, not just user dissatisfaction
- Adversarial robustness: Must remain operational under electronic warfare and cyber attacks
- Security isolation: Models must run on physically isolated classified networks
These requirements will drive Google’s progress in model safety and robustness, benefiting commercial products.
2. Global Chain Reactions
- Chinese model companies: May accelerate cooperation with domestic military and national security agencies
- European AI companies: Face the same “participate in defense or not” decision
- Open-source community: May introduce stricter license terms restricting military use
Actionable Advice
For AI Professionals
- Watch Google’s defense AI hiring — a high-growth area for the next 1-2 years
- AI safety/robustness roles will receive more resources due to defense demand
For Investors
- Defense AI market is shifting from “startup-led” to “Big Tech entry” — investment logic needs adjustment
- Watch Google defense AI suppliers and partners
For Developers
- If you care about AI ethics, watch open-source model license changes regarding military use
- Consider explicitly stating AI use restrictions in your projects