Google DeepMind resignations highlight concerns over Pentagon AI deals

Reuters reported that Google joined technology firms with a deal allowing the US Department of Defense to use its AI models for classified work.[3] The agreement allows the Pentagon to use Google’s AI for any lawful government purpose.[3] The BBC reported that the agreement says the models should not be used for domestic mass surveillance or autonomous weapons without appropriate human oversight and control.[2] Livemint reported that, in classified settings, Google has limited visibility and no clear veto over final uses.[1] Google says it reviews public-sector contracts for compliance with its acceptable-use policies and AI principles.[2]

Livemint reported that several DeepMind staff resigned in recent weeks over the military agreements.[1] Researchers who left said the deals expanded government applications beyond what they expected when they joined.[1] London-based research engineer Sofia Liguori left after three years at DeepMind; she said the technology automatically makes it easier to escalate the brutality of war and that she no longer believed it would have a net positive impact on society.[1] Andreas Kirsch also resigned earlier in the month, saying Google’s governance and oversight were not sufficient or meaningful enough for safe military AI deployment.[2]

Researchers were concerned that existing safeguards and human oversight might not be sufficient on air-gapped military networks.[1] More than 100 employees working on Google’s AI technology signed a letter to management objecting to Gemini being used to surveil American citizens or pilot autonomous weapons without human involvement.[4] The letter was sent to Google DeepMind chief scientist Jeff Dean.[4] Separately, nearly 50 OpenAI employees and 175 Google employees published a letter criticizing the Pentagon’s negotiating tactics.[4] Google staff told the BBC they feared speaking publicly about AI safety and warfare because employment contracts restricted public criticism; Google said employees have channels to raise concerns and policies intended to prevent retaliation.[2]