Andreas Kirsch | Trust is not Governance
Trust is not Governance
Google DeepMind's reported Pentagon contract shows why trust and safety culture can't substitute for real governance: independent oversight, transparency, accountability; and protected employee voice.
Originally published by UTAW.
Speaking in my personal capacity as a UTAW member, not on behalf of Google or DeepMind. This essay was written by me and the views expressed are my own.I am a member of UTAW (United Tech & Allied Workers), a branch of CWU (Communication Workers’ Union). The essay relates to issues reflected in the UTAW@Google campaign: AI principles and safeguards, transparency, military and surveillance uses of our work, moral refusal, whistleblower protection, and employees’ ability to raise ethical objections without retaliation. I wrote it outside working time, without Google resources or confidential information. I speak only for myself, not for Google, DeepMind, UTAW, CWU, Unite, or any colleagues.
Some stories survive long after they stop being true.
I joined DeepMind because it has taken AGI and ASI (artificial superintelligence)Artificial general intelligence is defined in many different ways, but an approximate definition is that “it matches or surpasses human capabilities across virtually all cognitive tasks” (Wikipedia). By doing so, it will allow for recursive self-improvement of AI systems and lead to artificial superintelligence, which would outperform human intelligence by wide margins on all tasks. seriously for longer than any other lab. It is a place where I hoped to work to maximise the likelihood of positive outcomes for humanity. Many people there share this belief, and they are more thoughtful and ethical than I.
For years, DeepMind has bet that a strong safety culture and good leadership built on trust are sufficient to withstand outside pressure. The “Pentagon” contract with the US Department of Defense, which Google reportedly signed on April 27th, is the most consequential test of this bet so far.Erin Woo, “Google Signs Classified AI Deal With Pentagon Amid Employee Opposition”, The Information, Apr. 27, 2026; Reuters, “Google signs classified AI deal with Pentagon, The Information reports”, Apr. 28, 2026. And one does not have to be categorically against military use of AI to feel troubled about it.
Given everything that is known, the bet has failed: good people do not make up for a lack of real governance. Like any frontier lab, Google DeepMind ought to have real governance that includes meaningful independent oversight with the authority to say no, transparency to employees and the public, and accountability when commercial or political pressure collides with stated principles. Employees should not be afraid to ask for this.
EFF: “Weasel Words”
We do not know the full language of the contract Google has signed, but the silence around what was signed and the reported contract language are revealing.
Google states that it is “committed to the private and public sector consensus that AI should not be used for domestic mass surveillance or autonomous weaponry without appropriate human oversight.”NBC News quoted this Google statement in its report on the Pentagon/Google agreement: David Ingram, “Pentagon inks deal with Google for AI services”, Apr. 28, 2026. Reuters reported substantially the same statement. But “should not” is not “must not,” and “appropriate human oversight” depends on who gets to define it. If these uses were truly prohibited, simply saying so would have taken fewer words.
According to The Information, the reported contract permits “any lawful government purpose,” requires Google to assist in adjusting safety settings and filters at the government’s request, and explicitly states that the terms do not allow Google to control or veto the government’s lawful operational decisions.Woo, The Information, Apr. 27, 2026. Reuters summarized the same reported terms, including “any lawful government purpose,” filter adjustments, “should not be used” language, and no Google right to control or veto lawful operational decisions. Once models run on classified networks, Google likely has little actual visibility into how the models are used.Google’s public Google Distributed Cloud air-gapped documentation says the system can “operate fully disconnected” from Google Cloud and that “consumption info is not visible in Google Cloud console”; its DISA page states that Google Distributed Cloud air-gapped and appliance offerings have DoD IL6 provisional authorization and can connect to SIPRNet.
Charlie Bullock, a lawyer and senior research fellow at the Institute for Law and AI, told The Information that Google’s phrasing “is not intended for, and should not be used for” is “not legally binding in any way.”Woo, The Information, Apr. 27, 2026. The Electronic Frontier Foundation has described OpenAI’s similar contract language as being full of “weasel words”:Corynne McSherry and Matthew Guariglia,...