A Manyfesto for Critical AI

Stanford, are you awake?

AI’s development accelerates while democratic institutions erode. These rapid global shifts are fueled by the very narratives exported from our own campus, and we recognize that this is not a neutral location. There is no Silicon Valley without Stanford, and no Silicon Valley imaginary without Stanford ideals. From the lean startup to the foundation model: the very grammar of disruption was codified in the rooms where we now sit. This university exports narratives as effectively as it generates capital, shaping how AI is built, absorbed by global institutions, and understood by the public. We write to assert that the power has flowed largely in one direction, and what it will take to redirect it.

Our group was born from a shared cognitive dissonance, the serendipitous crossing of paths between people from different fields and generations who could not, in good conscience, leave what was unfolding outside the gates of Stanford.

We hold that AI is not a neutral technology. It is socio-technical: co-produced with the institutions, labor arrangements, and political economies that surround it. Dominant AI discourse has been slow to absorb this fact. Treating AI as a purely technical phenomenon, a set of capabilities to be celebrated or feared in the abstract, mistakes its nature and disarms our response to it. To inquire critically into AI is to iteratively ask which interests it serves, which it forecloses, whose labor it conceals, and which alternatives were ruled out before the first model was trained.

We hold that AI’s trajectory is not given. The dominant narrative casts its development as a force of nature: inevitable, accelerating, beyond democratic reach. We reject the grammar of inevitability. As Terry Winograd observed, deciding is a computational activity, something that can ultimately be programmed; choice is the product of judgment, not calculation. The distinction between deciding and choosing lies between submitting to a system and exercising agency within and against it. We are here to defend that distinction, and to insist that the choices being made about AI, by firms, by states, by universities, are choices, not computations, and must answer and be held accountable to those they affect.

To be critical of AI is to be democratic and reasonable. We are guided not by resentment, nostalgia, or fear, but by the desire to scrutinize any unexamined powerful system. We ask who benefits and who bears the costs of this technology. We scrutinize biased assumptions embedded in the design, and dare to imagine alternative futures.

We believe the most important AI risks are strategically hidden from the mainstream. AI safety conversations pushed by big tech CEOs and major AI labs focus on “long-term” catastrophic risks, rogue superintelligence, and existential threats while dreaming for a technoutopia. We do not dismiss all long-horizon concerns, but believe some intentionally evade accountability and question. Existential risks fuel the distraction in which the true threats lie: censorship, oppression, and control. Monopolized platforms and algorithmic manipulation take away voices of justice. Data labeling, AI training, and extraction of raw materials comes at the cost of slave labor and the health of our local and global communities. Military and surveillance AI systems have insidiously gained the ability to identify, track, and target. We understand the material risks of AI are not speculative. These are happening now, and they disproportionately affect communities that were already marginalized far before the dawn of modern AI.

Our institutions and communities must create reform, and we must use history as guidance. Through contestation, resistance, and reimagination, we will name and dismantle the logics that advance colonial, authoritarian AI. We will reshape what it means to develop artificial intelligence without the entrenchment of bias and concentration of infrastructural power. In addition to embracing lived experiences, diverse voices within our civil society will allow us to envision a world in which technology can empower without exploitation.

What we are doing

We are constituting a community in which these questions can be asked without professional cost. Our membership runs from undergraduates to faculty, from computer science to law, communication, sociology, and the humanities, because the questions we ask do not respect disciplinary borders. We are prototyping forms of resistance, in research, in narrative, in design. We are creating space for tech workers, academics, and activists experiencing moral conflict about Silicon Valley’s and Stanford’s complicity to imagine and create alternatives.

We hold contradictions open rather than resolving them prematurely, because premature resolution is one of the ways power conceals itself. We do not have all the answers, but we are committed to pursuing the questions and acting together.

An invitation

This Manyfesto is an opening that invites scrutiny, revision, and the people who hold views we have not yet encountered. What is not optional, in this moment, is engagement. The choices being made now will shape institutions and lives for decades. Stanford will play a role in those choices, whether or not it reflects on them. We have decided to reflect.