◎

Behind the AI

Who funds, controls, and profits from the AI tools you use every day? Independent assessments. No corporate sponsors. Plain language.

◇

Google DeepMind

▲Serious Concern

World's largest data company. Quietly lifted AI weapons ban in 2025.

Key Facts

Company
Google DeepMind
Product
Gemini
Grade
Serious Concern
Overall Score
2.4 / 5
Last Updated
2026-03-22
Military Status
Active
Last Reviewed2026-03-22
Company ContactedNot yet contacted
MethodEmail to [email protected]
ResponseNo response received
Data Confidence: HighAssessed by: Behind the AI Research Team
Founded2010 / 2023
HQLondon & Mountain View
StructureDivision of Alphabet ($2T+)
CEODemis Hassabis / Sundar Pichai
ValuationPart of Alphabet ($2T+)
RevenueGoogle Cloud $40B+/yr
Core Identity

Breakthroughs in biology (AlphaFold) and health (MedGemma). But research division increasingly subordinate to Alphabet's commercial and government ambitions.

Military & Government

Pentagon cloud contracts. Lifted internal weapons/surveillance ban Feb 2025. Project Maven (2018). Project Nimbus ($1.2B contract; over 70% of revenue from Israeli Ministry of Defense; could reach $3.3B by 2027; leaked documents revealed secret provision to alert Israel if foreign court orders data handover). $200M CDAO contract (July 2025). Gemini deployed on GenAI.mil (3M personnel). $9B JWCC cloud contract share. Agent Designer access granted to Pentagon (Feb 2026). Classified Gemini deployment negotiations ongoing. 100+ DeepMind employees signed letter opposing military work. 200+ employees signed open letter supporting Anthropic.

Assessment by Pillar

Data Practices

Free Gemini trains on conversations. Google already has Search, Gmail, Maps, YouTube, Android data. Gemini adds conversational data to that profile.

Funding Transparency

Public company. Financials fully disclosed. But core business is surveillance advertising.

Military & Gov't

Active Pentagon contracts. Lifted weapons ban without public announcement. Project Nimbus provides AI to Israeli military.

Model Transparency

Some research published. AI Principles exist but contain loopholes. Original DeepMind Ethics Board (2014 acquisition condition) likely never convened at all. Google's ATEAC dissolved before its first meeting in 2019. Current Internal Review Committee (IRC) meets biweekly.

User Rights

Paid tiers opt out of training. History deletion available. Enterprise offers Canadian data residency (Montreal). But Google's broader data collection continues.

Key Findings

RISK

Using free Gemini adds your conversations to the data profile Google already builds from your Search, Gmail, Maps, YouTube, and Android activity.

RISK

Google quietly lifted its internal AI weapons ban in Feb 2025 without public announcement. The AI Principles published after the 2018 Maven controversy have been steadily loosened.

CONTEXT

100+ DeepMind employees are actively opposing military contracts from inside the company. This internal dissent suggests the research team's values diverge from corporate direction.

What They Say vs. What Happened

What they said

Committed to ethical AI. Created a dedicated Ethical AI team.

What happened

Fired Timnit Gebru (Dec 2020) and Margaret Mitchell over the "Stochastic Parrots" paper. 2,600+ Googlers signed a protest letter. Quietly lifted AI weapons ban (2025). Emissions up 51% since 2019.

Bright spot

Best PUE (1.09), only company doing hourly carbon-free energy matching (66%), published per-query methodology (Google Cloud Blog Aug 2025).

Source

Firing of Ethical AI team leads and emissions data

Karen Hao, MIT Tech Review Dec 2020; Google 2025 Environmental Report

The Pattern

↓Ethics commitment announced
↓Ethics team hired
↓Team finds real problems
↓Problems threaten revenue
↓Team defunded, fired, or redirected
↓New commitment announced
↺Repeat

“They always do just enough to be able to put the press release out.”

— Hany Farid, UC Berkeley

The Law Is Catching Up

●
EU AI Act transparency obligations ↗Enforcement begins August 2, 2026
●
73 new US state AI laws in 2025 ↗Rapid state-level legislative action across the US
●
AI Whistleblower Protection ActPrompted by OpenAI NDA revelations threatening departing employees with equity loss
What to look for in an alternative

A provider whose core business is not data collection for advertising, and whose military/government policies are publicly documented rather than internally revised.

Who Funds Them

Alphabet (self-funded) World's largest ad companyBillions/year

Recent News

Feb 2025: Lifted internal AI weapons/surveillance ban
Mar 2026: 100+ DeepMind employees oppose Pentagon work
Mar 2026: Jeff Dean filed amicus brief supporting Anthropic
Mar 2026: 900+ Google and OpenAI employees signed letter supporting Anthropic

Frequently Asked Questions — Google DeepMind

Does Google Gemini train on my conversations?

Free Gemini trains on your conversations. Google One AI Premium ($20/month) opts out of training. Google still collects data from Search, Gmail, Maps, YouTube, and Android regardless of your Gemini tier.

Does Google have military AI contracts?

Google has active Pentagon cloud contracts and lifted its internal ban on AI for weapons and surveillance in February 2025. It also has Project Nimbus, a contract with the Israeli military.

Sources

  1. [1]Lifted internal weapons/surveillance ban Feb 2025— MLQ AI News, Mar 2026
  2. [2]Project Maven controversy 2018— Multiple / Public record, 2018
  3. [3]Project Nimbus: Israeli military contract— TIME, Aug 2024
  4. [4]100+ DeepMind employees signed letter opposing military work— NYT (via Web And IT News), Feb 27, 2026
  5. [5]200+ Google employees signed open letter supporting Anthropic— The Week, Feb 27, 2026
  6. [6]DeepMind acquired 2014 with promise tech wouldn't be used for military— TIME, Aug 2024
  7. [7]Original DeepMind Ethics Board likely never convened. ATEAC dissolved before first meeting (2019). Current IRC meets biweekly.— TIME, Aug 2024

Related Assessments

Disclaimer

Assessments reflect publicly available information and the published methodology of the Behind the AI Research Team. Grades represent analytical assessments derived from the published scoring framework, not statements of fact about internal company operations. If you believe any claim is inaccurate, contact [email protected] with the specific claim and your evidence.