Google employees urge CEO to reject Pentagon AI deal

On April 27, more than 600 Google employees urged CEO Sundar Pichai to reject a proposed Pentagon deal that would let the company’s Gemini AI be used in classified military operations.

The signatories included staff from DeepMind, Cloud, and other teams, including more than 20 senior leaders.

Organizers said Google proposed safeguards to block domestic mass surveillance and autonomous weapons without human control.

Employees said those safeguards would be difficult to enforce because Pentagon policy limits external oversight of its AI systems.

The dispute comes as Google seeks to extend an existing Defense Department contract into classified work after ending Project Maven in 2018.

🔗 Source: The Economic Times

🧠 Food for thought

Implications, context, and why it matters.

Google changes its AI rules as Pentagon demand grows

Google last year removed a section from its AI Principles that had promised to avoid potentially harmful uses such as weapons. The company adopted those rules after employee protests over Project Maven. The world had changed and US tech companies had a duty to help the country defend itself, said Demis Hassabis, CEO of Google DeepMind 1.
The change came as Microsoft already had deals to provide AI in classified settings, while OpenAI announced a revised Pentagon agreement in February 2.
The Pentagon has clashed with Anthropic after labeling the startup a “supply chain risk” during a dispute over guardrails, or limits on military AI use. That fight pushed the Pentagon to seek other partners and opened a path for Google as it discusses a classified Gemini deal 1.

The dispute could divide AI companies over defense work

The argument turns on whether tech companies can keep ethical red lines in military AI used in “air-gapped” classified systems, or secure computers cut off from the public internet. Google employees argue it would be hard to track use or enforce limits in those systems 3.
If Google walks away, rivals may take more of the defense AI market. That could split the industry between firms that accept military terms and firms that focus on commercial work or ethical limits 3.
The broader issue is whether private companies should set usage rules during wartime or during preparations for war. The Pentagon has pushed back hard on that idea 3.