René Mayrhofer, the engineer who led Google‘s Android platform safety workforce as a director earlier than transferring right into a principal engineer function, has resigned over the corporate’s resolution to let the Pentagon use its Gemini AI for labeled navy work. His farewell be aware pulls no punches—he informed colleagues that management “has misplaced its ethical compass,” the road he used to title the letter itself. The be aware was first obtained and reported by Enterprise Insider.Mayrhofer dated the be aware Might 18 and circulated it internally earlier than publishing it on his private weblog, the place he says it “unfold far wider than he anticipated.” His core cause is blunt: as a self-described pacifist, he will not work even not directly for a US navy he sees as engaged in offensive warfare. “Proactively harming individuals is just not one thing that I can or will probably be concerned with,” he wrote.
The Pentagon AI deal that drove the resignation
The set off was Google’s late-April settlement with the Division of Protection, not too long ago renamed the Division of Warfare below the Trump administration. The contract clears the Pentagon to run Google’s Gemini mannequin on labeled methods for “any lawful goal”—the identical open-ended wording OpenAI and Elon Musk’s xAI accepted in their very own navy offers.That phrase is precisely what set Mayrhofer off. Google, he wrote, is “signing offers with the US Ministry of Warfare,” the place “any lawful goal” by the present authorities had already been proven, repeatedly, to interrupt worldwide legislation. He confirmed the letter’s authenticity to Enterprise Insider and stated the choices have been made totally on the high, with out inner debate. “None of that is being debated or communicated throughout the firm,” he wrote—a pointy criticism from somebody who’d as soon as sat in Google’s administration chain himself.Mayrhofer led the Android platform safety workforce from Mountain View till round 2019, then shifted to a part-time, strategic function after relocating his household to Austria. Over a yr in the past he formally moved from Director Engineering to a principal software program engineer submit on the individual-contributor monitor. He wasn’t a sitting director when he resigned—a element he pointedly corrected.
Why surveillance fears made it private
For Mayrhofer, the stakes aren’t summary. He is additionally a professor at Johannes Kepler College Linz in Austria, and he fears the “any lawful goal” clause might stretch to cowl mass surveillance of EU residents, himself included. He pointed to the US authorities turning “hostile” towards European lecturers and warned that Google’s AI merchandise “will seemingly be used instantly towards me and mine.”The local weather angle feeds his broader learn that Google’s priorities have quietly shifted. In the identical be aware he wrote that the corporate had deserted its carbon-neutral purpose to feed the vitality urge for food of its AI fashions—a call he frames as a part of the identical top-down drift, not a co-equal cause for leaving.His departure can also be a measure of how far Google has moved. He signed the 2018 open letter that pushed the corporate to drop an earlier Pentagon contract, again when Sundar Pichai’s AI ideas explicitly dominated out weapons and surveillance instruments. These clauses have been stripped out in February 2025. The “Do not be evil” motto had pale lengthy earlier than.
A wider revolt inside Google over navy AI
Mayrhofer is way from the one one uneasy. Greater than 560 staff signed an open letter urging Pichai to stroll away from the talks, warning the know-how could possibly be utilized in “inhumane or extraordinarily dangerous methods.” Andreas Kirsch, a analysis scientist at Google DeepMind, informed Enterprise Insider in April he was “extremely ashamed” of the choice.Google did not reply to Enterprise Insider’s request for remark, although a spokesperson beforehand stated the corporate was “proud” to help nationwide safety and maintained its AI should not energy home mass surveillance or autonomous weapons with out human oversight.Mayrhofer, who concedes his tenure made it simpler to go away than it could be for colleagues who rely on the paycheck, plans to serve out his discover till August. He’ll cease touching any AI work tied to the deal instantly. “I’m fairly unhappy that it needed to come to this,” he wrote, “and desperately hope Google administration re-discovers its ethical compass.”
The total farewell be aware, as printed on Mayrhofer’s weblog:
When Google provided me the job of Director of Android Platform Safety in 2017, it was not possible to refuse. Sure, Trump was already president—my household and I had qualms—however he appeared contained, even ineffective. Extra importantly, Google was a distinct firm 9 years in the past. Android was open supply first and had simply surpassed 2 billion customers. I would been learning its safety from the surface since 2009, and it was (and nonetheless is! ) probably the most thrilling end-user going through working system to work on. Nevertheless, whereas the supply code was at all times public, getting direct contact to the interior Android workforce had been extremely tough; making an attempt to debate new concepts for safety mitigations or architectures supporting upcoming fields like cellular digital ID was a irritating train for lecturers and trade researchers exterior Google.Getting the possibility to steer on the within, on probably the most broadly used Linux primarily based, (principally) open supply working system on the planet, was an unimaginable probability. I’m nonetheless grateful for the preliminary provide, particularly to Dave Kleidermacher and Nick Kralevich for his or her belief in me, and the welcoming ambiance from day one. Google was the place to be to getting issues finished on a worldwide scale, The tradition was clear and open to numerous discourse, and from the beginning it was made clear that, as Googlers, we weren’t solely welcome however anticipated to carry our personal identification and values into the job. As an educational and tenured professor of pc safety, engaged on Android inside Google was actually probably the most interesting place in the entire of the Silicon Valley – the one which finest matched the spirit of academia and my very own moral ideas to work for the general public good.Whereas I used to be by no means actually concerned with the cloud aspect of issues, on the corporate degree, the purpose was nonetheless to turn out to be utterly carbon-neutral, and contracts with the Pentagon have been canceled after staff spoke up towards them (I signed the 2018 open letter). The AI ideas printed by Sundar Pichai in 2018 acknowledged very clearly that “AI purposes we is not going to pursue: … 2. Weapons or different applied sciences whose principal goal or implementation is to trigger or instantly facilitate damage to individuals. 3. Applied sciences that collect or use info for surveillance violating internationally accepted norms. 4. Applied sciences whose goal contravenes broadly accepted ideas of worldwide legislation and human rights.” Many pc scientists and software program engineers needed to work at Google, and I heard each hearty congrats and fierce jealousy after I talked about the job provide to colleagues earlier than relocating to Mountain View.Then there have been the individuals. Larry and Sergey have been nonetheless answering some powerful management questions each week, and “Do not Be Evil” wasn’t only a slogan of often-referenced Googliness—it was a north star for groups making exhausting calls. My rapid workforce—Android Safety, the defenders of Billions of customers—has the motto to “make issues so safe that we ourselves cannot break them, whether or not the machine prices $1000 or $100, or the person is a celeb or a refugee”. It was at all times about doing proper for our customers, and defending their pursuits first (often even towards enterprise pursuits of different Google apps and companies). I met probably the most superb consultants inside my first months of becoming a member of, together with Android legends like Dianne Hackborn. Everyone was pleasant, joyful to provide time to newcomers, to share their data in regards to the know-how in addition to in regards to the inner processes. And all people was devoted to do proper by the worldwide inhabitants—thanks lots to all of you for that onerous work! I’m nonetheless extremely happy with a lot of our achievements, most of which required transferring different ecosystem stakeholders over lengthy intervals of time. Making full machine encryption the Android 10 default even for the most affordable of gadgets moved the world ahead. Enabling end-to-end encrypted Android backup quietly whereas the discussions targeted on Apple outlined a de facto state-of-the-art that also holds robust in present legislation enforcement vs. person privateness discussions. Insider Assault Resistance, ARM MTE, privacy-first digital credentials, and plenty of different issues have been solely doable as a result of we pulled collectively to make our customers safer—together with towards a doubtlessly malicious sub-part of Google itself.Sadly, occasions have modified. Google administration has quietly deserted its objectives to turn out to be carbon-neutral due to the AI mannequin vitality utilization. Worse, Google administration is now signing offers with the US Ministry of Warfare—the place “any lawful goal” by the present US authorities has already been repeatedly demonstrated to be in violation of worldwide legal guidelines. None of that is being debated or communicated throughout the firm. It’s simply determined by top-level administration (I used to be a part of the administration chain earlier than, and I hadn’t heard of any of those adjustments via inner channels). With my ethical and moral ideas, I can not—explicitly or implicitly, instantly or transitively—help the present and ongoing actions of the “Most lethality, not tepid legality” US Ministry of Warfare. Given Google’s top-level administration course and up to date doubling-down, this sadly leaves me with the one option to resign.On the one hand, this resolution has been extremely exhausting to make. I’ll miss the individuals, all of you who’re nonetheless making an attempt to do good for the remainder of the planet. I’ll miss the alternatives to have an effect on constructive change. I’ll miss the sensible engineers and technically targeted decision-making. I’ll miss the innocent post-mortems and the general, very mature tradition on coping with failures.Then again, this resolution has been simple as a result of it has turn out to be unavoidable. I’m a pacifist, and have way back determined that I can’t personally work for militaries participating in offensive warfare (strictly defensive motion is considerably completely different). Proactively harming individuals is just not one thing that I can or will probably be concerned with. I’m additionally a European educational. Meaning the present US authorities has turn out to be hostile to me, and “any lawful goal” on this sense will completely embody mass surveillance of EU residents. This deal implies that Google (AI) merchandise will seemingly be used instantly towards me and mine. On this current setting, I do not see how I couldn’t resign.My present contract provides a discover interval of three months beginning with the final day of the month through which the resignation is tendered. Meaning I am going to nonetheless be round (in my restricted time dedication) and reachable via inner channels till 2026-08-31, wrapping up or passing on a few of my ongoing tasks—however I’ll instantly disconnect from any work on AI methods which may fall below this cope with the DoW (not that I’m conscious of getting been concerned up to now). Afterward, I needs to be simple to achieve externally via a number of channels. I’ll proceed to work on end-to-end encrypted, resilient communication and storage protocols, privacy-preserving digital identification, embedded methods safety, working methods and provide chain safety, and associated matters. One intersection level of those matters is clearly nonetheless Android (significantly AOSP) safety and privateness.I’m fairly unhappy that it needed to come to this, and desperately hope Google administration re-discovers its ethical compass. Till then, I am going to miss y’all.