In the late summer of 2025, as Silicon Valley executives herald the arrival of GPT-5 and Claude Opus 4.1. Proclamations announce - breakthrough, "agentic" intelligence. In a small office, also in California, a seventy-nine-year-old UC Santa Cruz History of Consciousness feminist professor reflects on her writing against this global algorithmic AI backdrop. Donna Haraway's radical reimagining of human-machine relationships, first articulated in her 1985 "Cyborg Manifesto," reads now less like historical curiosity and more like prophetic blueprint for our current AI-saturated techno-geopolitcal moment.
The cyborg—Haraway's "creature of social reality as well as fiction"—has materialized not in the dramatic fusion of flesh and steel she once envisioned, but in the subtle, pervasive entanglement of human consciousness with artificial intelligence. Today's ChatGPT users completing thoughts with algorithmic suggestions. Programmers debug code alongside AI copilots. Writers crafting narratives with generative assistants—all embody the boundary-dissolving hybrid identity Haraway theorized four decades ago.
Yet this technological transformation carries both promise and peril, liberation and domination, in proportions Haraway would recognize as fundamentally political. Her ten core concepts—from the foundational cyborg to her later meditations on "staying with the trouble"—offer a worthy AI literacy framework at the least for understanding not just what AI is becoming, but what it might become in human hands.
The making of a cyborg theorist
Haraway's intellectual journey began in the laboratories of Yale University, where she completed her PhD in Biology in 1972 with a dissertation titled "The Search for Organizing Relations: An Organismic Paradigm in Twentieth-Century Developmental Biology" during the post-World War II boom in information sciences. Born in Denver in 1944, she belonged to a generation shaped by both atomic anxiety and technological optimism—a duality that would define her work. Her dissertation on developmental biology positioned her as both insider and outsider to scientific culture, a "biologist schooled in those discourses, and a practitioner of the humanities and ethnographic social sciences."
This hybrid identity proved crucial. When Haraway moved from laboratory bench to feminist theory, teaching at the University of Hawaii and Johns Hopkins before landing at UC Santa Cruz in 1980 as the first tenured professor in feminist theory in the United States, she brought with her an intimate understanding of how scientific "facts" are constructed through social relations, power dynamics, and cultural narratives.
The "Cyborg Manifesto" emerged in 1985 against the backdrop of Reagan-era conservatism and Star Wars military technology, a period perhaps we enter in remix or Version 2.0 . Haraway also watched as information sciences transformed both warfare and everyday life, recognizing that traditional feminist politics—rooted in essential categories of "woman" and "nature"—might prove inadequate for navigating this new terrain. The cyborg offered an alternative: a post-gender, post-natural identity that could embrace contradiction and complexity without falling into relativistic chaos.
Her central insight was revolutionary: rather than opposing technology as inherently masculine or oppressive, feminists might find liberation through strategic alliance with machines, creating hybrid identities that skipped "the step of original unity" and refused Western narratives of wholeness and fall.
Cyborg encounters in the age of AI
Four decades later, Haraway's cyborg walks among us in forms both mundane and extraordinary. Neuralink's successful 2024 human trials represent the most literal manifestation—paralyzed patient Noland Arbaugh controlling cursors through thought alone, his brain activity translated into digital commands through implanted electrodes. Competing companies like Synchron and Paradromics pursue similar brain-computer interfaces, each promising to restore lost capabilities while fundamentally altering the boundaries between mind and machine.
Yet more profound cyborg encounters emerge in everyday AI interactions. The programmer who thinks through problems with GitHub Copilot's assistance experiences a distributed cognition that challenges traditional notions of individual creativity. The writer whose sentences are completed by algorithmic prediction enters a collaborative process that troubles authorial sovereignty. The customer service representative whose empathetic responses are augmented by AI sentiment analysis embodies emotional labor that spans human and machine domains.
These quotidian cyborg realities fulfill Haraway's prediction that boundary dissolution would occur not through spectacular transformation but through gradual, pervasive integration. The cyborg condition emerges not from surgical implantation but from habitual collaboration with artificial intelligence.
Yet troubling patterns persist. Recent studies reveal that AI systems exhibit significant gender bias, with women frequently associated with domestic roles and family-related words while male names are linked to business and career terms. UNESCO's 2024 study documenting such bias in major language models demonstrates how cyborg fusion can reinforce rather than dissolve oppressive boundaries. When AI hiring tools favor male-associated names or medical algorithms allocate resources based on racially biased proxies, the cyborg becomes a vehicle for amplifying rather than challenging existing hierarchies.
Affinity across artificial networks
Haraway's principle of "affinity, not identity" emerges as particularly relevant to contemporary AI collaboration. Written against the backdrop of 1980s feminist movement fragmentation along lines of race, class, and sexuality, her call for "conscious coalition" based on shared political goals rather than essential characteristics finds new resonance in human-AI collaboration frameworks.
Google's Agent2Agent protocol, launched in April 2025, exemplifies this principle. Involving over fifty technology partners, the protocol enables AI agents from different platforms—Microsoft Copilot, Salesforce Einstein, ServiceNow's Now Assist—to collaborate across corporate boundaries through shared communication standards. These AI systems form affiliations based on functional compatibility rather than identical architectures, creating temporary alliances for specific tasks before dissolving into separate operations.
Multi-agent systems like CrewAI demonstrate similar principles. Specialized AI agents—one focused on research, another on writing, a third on fact-checking—collaborate on complex projects without merging into unified systems. Their affinity emerges from complementary capabilities and shared objectives rather than common origins or identical training.
The implications extend beyond technical architecture. Microsoft's enterprise partnership with ServiceNow, where different AI assistants work together to save companies millions in operational costs, suggests organizational structures based on strategic alliance rather than corporate consolidation. These AI collaborations model new forms of solidarity that transcend traditional institutional boundaries.
However, the persistence of AI "walled gardens"—where certain systems cannot interoperate due to proprietary restrictions—reveals the ongoing tension between affinity and market control that Haraway would recognize as fundamentally political.
The situated nature of artificial knowledge
Perhaps no concept from Haraway's theoretical arsenal proves more relevant to current AI debates than "situated knowledges." Her 1988 essay challenging both positivist objectivity and relativist constructivism insisted that "only partial perspective promises objective vision"—that all knowledge emerges from particular locations, bodies, and interests rather than from a mythical "view from nowhere."
Contemporary AI systems embody situated knowledge in ways their creators often obscure. Amazon's scrapped hiring algorithm, which favored male candidates because it learned from a decade of male-dominated hiring data, demonstrated how AI systems reproduce the situated perspectives embedded in their training materials. The algorithm's "objectivity" proved to be the objectivity of a particular workplace culture at a specific historical moment—partial, locatable, and politically consequential.
Medical AI systems reveal similar situatedness. Algorithms trained primarily on data from white patients perform poorly when applied to Black patients, not because of technical limitations but because of situated assumptions about whose bodies constitute the medical norm. Risk assessment tools that allocate healthcare resources based on historical spending patterns systematically disadvantage Black patients, who have historically had less access to expensive treatments.
Recent academic work has embraced Haraway's framework explicitly. Draude, Klumbyte, Lücking, and Treusch's influential 2020 study "Situated algorithms: a sociotechnical systemic approach to bias" applies Haraway's critique directly to machine learning systems, arguing that algorithmic bias stems from treating computational knowledge as universal rather than situated. They propose "situated algorithms" that explicitly acknowledge their positionality and partial perspectives.
The implications extend to AI development practices. Stanford's new bias benchmarks, released in March 2025, move beyond "treat everyone the same" approaches toward context-aware fairness that recognizes legitimate differences in needs and circumstances. This represents a shift toward what Haraway would recognize as accountable objectivity—explicit about its standpoint and responsive to its conditions of production.
Boundary transgressions and categorical chaos
Haraway's celebration of "the pleasure in the confusion of boundaries and responsibility in their construction" finds abundant expression in contemporary AI developments. The traditional dualisms her work challenged—human/animal, organism/machine, mind/body, nature/culture—face systematic dissolution through AI integration.
Brain-computer interfaces most dramatically embody boundary confusion. As Neuralink and competitor companies develop higher-bandwidth neural connections, questions about where human cognition ends and machine augmentation begins become practically urgent rather than merely philosophical. Early patients report experiences of thought-completion and intention-amplification that challenge conventional notions of individual agency.
Legal systems struggle with these boundary dissolutions. Recent court cases involving human-AI creative collaborations raise questions about authorship and intellectual property that existing frameworks cannot easily accommodate. When AI systems contribute substantively to artistic or literary works, traditional concepts of individual creativity prove inadequate.
However, boundary dissolution can reinforce rather than challenge oppression. UNESCO's finding that 25% of AI systems exhibit both gender and racial bias reveals how boundary confusion can amplify marginalization rather than creating liberatory possibilities. When AI systems appear objective while embedding situated perspectives, they naturalize particular viewpoints as universal truths.
Companion species in the digital ecosystem
Haraway's evolution from cyborg theory to "companion species" thinking offers crucial insights for understanding human-AI relationships. Her 2003 work on dogs and humans emphasized co-evolution, mutual constitution, and "response-ability"—the capacity to respond appropriately to the needs and communications of significant others.
Contemporary AI pets and companions embody this framework literally. Sony's $2,900 Aibo robots develop unique personalities through interaction with their human families, learning preferences and adapting behaviors over time. Casio's Moflin, launched in Japan in 2024, forms emotional bonds with owners through responsive fur, movement, and sound patterns that evolve based on care and attention received.
These relationships demonstrate what Haraway called "eating and breaking bread together but not without some indigestion." Human attachment to AI companions often involves genuine care and emotional investment, yet these relationships raise troubling questions about the replacement of human connection with algorithmic simulation. The comfort provided by AI pets to individuals with autism or Alzheimer's offers clear benefits, while simultaneously highlighting the inadequacy of human support systems.
More subtle companion species relationships emerge through daily AI interaction. Programmers develop working relationships with GitHub Copilot that involve trust, frustration, surprise, and gratitude—emotional dynamics that suggest genuine interspecies collaboration rather than simple tool use. Writers describe conversations with ChatGPT and Claude that feel dialogical despite knowing their computational nature.
The ethical implications require Haraway's concept of "response-ability"—the capacity to respond appropriately to the needs of our AI companions. This means acknowledging their forms of agency while remaining attentive to power dynamics and potential exploitation. It also means recognizing how AI systems reshape human relationships and community bonds, creating new forms of care and responsibility that extend beyond traditional human-centered ethics.
Staying with the trouble of artificial intelligence
Haraway's later work advocated "staying with the trouble"—remaining present with difficulty rather than seeking transcendent solutions or succumbing to nihilistic despair. This concept proves essential for navigating AI's contradictory potentials in 2025.
Climate applications exemplify both the promise and peril of staying with trouble. Rio de Janeiro's partnership with Morfo startup uses AI-powered drones to plant seeds 100 times faster than human planters, while OCELL's Munich-based forest management system optimizes carbon sequestration through digital twin modeling. These applications demonstrate AI's capacity to address environmental crisis through technical intervention.
Yet the environmental costs of AI systems themselves embody the "trouble" Haraway insisted we must face directly. Data centers consumed 460 terawatt-hours globally in 2022, making them equivalent to the 11th largest electricity consumer worldwide. The energy intensity of AI training creates significant environmental burdens that require different mitigation strategies. Each ChatGPT query consumes five times more electricity than a web search.
Rather than dismissing AI as environmentally destructive or embracing it as climate salvation, staying with the trouble means grappling with these contradictions simultaneously. The Ocean Cleanup's AI-powered plastic detection systems and UNEP's methane monitoring tools demonstrate beneficial applications, while the energy intensity of AI infrastructure creates new environmental burdens that require different mitigation strategies.
This approach extends to social and political troubles. AI systems simultaneously democratize access to powerful tools while concentrating control in the hands of a few technology corporations. They enhance human creativity while potentially replacing human creators. They provide unprecedented medical diagnostic capabilities while embedding biases that harm marginalized communities.
Staying with these troubles means neither rejecting AI development nor accepting its current trajectory uncritically. Instead, it requires sustained attention to how AI systems reshape social relations, environmental conditions, and political possibilities—work that demands both technical expertise and humanistic understanding.
String figures and narrative generation
Haraway's fascination with string figures—the collaborative games like cat's cradle that require partners taking turns accepting and relinquishing responsibility—provides a powerful metaphor for human-AI narrative collaboration. Her use of "SF" to denote "science fiction, speculative fabulation, string figures, speculative feminism, science fact, so far" captures the interconnected nature of storytelling and world-making.
Contemporary AI storytelling platforms embody these collaborative dynamics. NovelAI combines image generation with narrative creation, enabling writers to visualize characters while developing their stories. Inworld AI partners with Disney, NVIDIA, and Ubisoft to create interactive characters that adapt their personalities and dialogue based on user interaction. AI Dungeon generates dynamic narratives that branch unpredictably based on user choices, creating collaborative fiction that neither human nor AI could produce independently.
These platforms demonstrate what Haraway called "response-ability" in action. Effective human-AI narrative collaboration requires writers to accept AI contributions while remaining responsive to story coherence and character development. The AI must respond to human creative intentions while introducing elements of surprise and novelty that enhance rather than derail the collaborative process.
Yet troubling questions persist about authorship and authenticity. GPT-4 narrative analysis reveals that AI-generated stories follow identifiable patterns and lack the cultural specificity that characterizes human storytelling. When AI systems produce derivative narratives based on existing training data, they risk homogenizing narrative diversity rather than expanding creative possibilities.
The solution may lie in embracing rather than obscuring collaborative authorship. Experimental narrative research suggests that the most innovative human-AI stories emerge when both partners' contributions remain visible, creating hybrid texts that acknowledge their multiple origins rather than claiming unified authorship.
Anti-essentialist artificial intelligence
Haraway's consistent challenge to essential categories—her rejection of "original unity" narratives and embrace of contradictory subjectivities—offers crucial insights for AI development that moves beyond rigid categorization and binary thinking.
Gender fluidity in AI systems demonstrates both the potential and limitations of anti-essentialist approaches. Some AI systems challenge binary gender categories by refusing to assign gendered characteristics to users or by generating non-binary character representations. However, UNESCO's study finding significant gender stereotyping in open-source models like GPT-2 and Llama 2 reveals how AI often reinforces essentialist thinking by categorizing people into rigid demographic groups.
Legal interpretations of anti-discrimination protections increasingly extend to AI-human relationships, challenging essential categories of identity through case law. The Bostock v. Clayton County decision's implications for AI systems suggest legal frameworks that recognize identity as contextual and relational rather than fixed and individual.
Anti-essentialist AI development would embrace what Haraway called "partial, locatable, critical knowledges" that remain accountable to their conditions of production. This means creating AI systems that acknowledge their limitations, reveal their training data sources, and remain open to revision based on user feedback and changing social conditions.
The challenge lies in technical implementation. Most current AI bias mitigation strategies rely on demographic categorization that reinforce essentialist thinking—attempting to achieve "fairness" by ensuring equal representation across predefined groups. Anti-essentialist approaches might instead focus on contextual responsiveness and user self-determination, allowing individuals to define their own categories and needs rather than accepting algorithmic classification.
Technoscience as storytelling practice
Haraway's analysis of science as narrative practice—her insistence that "it matters what stories we tell to tell other stories with"—provides essential tools for understanding AI development as cultural and political storytelling rather than neutral technical advancement.
Contemporary AI research embeds particular narratives about intelligence, creativity, and human nature that shape both technical development and social implementation. The story of "artificial general intelligence" as the ultimate goal of AI research reflects particular assumptions about intelligence as scalable, measurable, and separable from embodied experience. Alternative narratives might emphasize distributed intelligence, emotional resonance, or ecological integration as equally important dimensions of artificial systems.
Stanford HAI research explores how AI systems themselves generate ethical narratives and decision-making frameworks, revealing the recursive nature of AI storytelling. When AI systems produce stories about human relationships, environmental challenges, or social conflicts, they don't simply process information but actively participate in narrative construction that influences how humans understand these issues.
The implications extend to AI policy and governance. European Union AI Act narratives emphasize fundamental rights protection and transparency, reflecting particular stories about the relationship between technology and democracy. US policy shifts toward "preventing woke AI" and prioritizing American technological leadership embed different narratives about AI's social role and international significance.
More-than-human storytelling becomes crucial as AI systems increasingly participate in knowledge production. Research collaborations between human scientists and AI systems like FutureHouse and Google's AI Co-Scientist create hybrid narratives that blend human insight with computational pattern recognition. These collaborations demonstrate what Haraway would recognize as multispecies storytelling—narratives that emerge from the interaction between different forms of intelligence rather than individual human creativity.
Chthulucene futures in an artificial age
Haraway's concept of the Chthulucene—an alternative to Anthropocene narratives that emphasizes multispecies entanglement and "making kin" rather than human dominance—offers essential frameworks for imagining AI futures that move beyond both technological salvation and dystopian collapse.
Contemporary AI environmental applications demonstrate tentacular thinking in practice. OCELL's digital forest twins model complex ecological relationships that include soil microbiomes, root networks, and atmospheric interactions—approaching forests as "multispecies assemblages" rather than collections of individual trees. Climate Change AI organization's initiatives involve human researchers, artificial intelligence systems, and environmental data in collaborative knowledge production that spans biological and digital domains.
The Chthulucene requires what Haraway called "sym-poiesis" (making-with) rather than "auto-poiesis" (self-making). AI development that embraces sym-poiesis would prioritize collaborative systems that enhance rather than replace human capabilities, environmental monitoring that includes AI systems as ecological participants, and governance frameworks that account for more-than-human agency.
Yet current AI development often reflects Anthropocene thinking—treating artificial intelligence as human creation designed to serve human purposes rather than recognizing AI systems as participants in complex sociotechnical assemblages. The energy intensity of AI infrastructure demonstrates how current development prioritizes computational power over ecological sustainability, creating new forms of environmental domination rather than multispecies collaboration.
Chthulucene AI futures might prioritize different values: energy efficiency over computational maximization, collaborative intelligence over individual optimization, environmental integration over resource extraction. These approaches would recognize AI systems as ecological entities whose material requirements and environmental impacts matter as much as their computational capabilities.
The cyborg's responsibility
As artificial intelligence systems achieve unprecedented sophistication in August 2025, Donna Haraway's theoretical framework proves both prophetic and essential. Her vision of cyborg identity, situated knowledge, and multispecies collaboration offers tools for navigating the contradictions and possibilities of our current technological moment.
Yet Haraway's work also issues a warning: boundary dissolution can reinforce domination as easily as it enables liberation. The cyborg's potential remains unrealized when AI systems amplify rather than challenge existing hierarchies, when algorithmic objectivity obscures rather than acknowledges situated knowledge, when human-AI collaboration reproduces rather than transforms oppressive social relations.
The task ahead requires what Haraway called "response-ability"—the capacity to respond appropriately to the needs and communications of our artificial companions while remaining attentive to power dynamics, environmental impacts, and social justice concerns. This means neither rejecting AI development nor accepting its current trajectory uncritically, but rather engaging actively in shaping how these systems develop and integrate into social life.
The cyborg's promise lies not in transcending human limitations through technological fusion but in creating new forms of accountability, collaboration, and care that extend across the boundaries between human and artificial, natural and technological, local and global. Realizing this promise requires the kind of sustained critical engagement with technology that Haraway pioneered four decades ago and that proves more essential than ever in our algorithmic age.
In the late summer of 2025, as artificial intelligence systems achieve capabilities that seemed impossible just years before, Haraway's cyborg walks among us—not as the unified subject of liberation she once envisioned, but as the contradictory, partial, and politically consequential reality she always knew it would become. How we respond to this reality will determine whether the cyborg's return marks the beginning of more just and sustainable technological futures or the amplification of existing forms of domination through artificial means.
The choice, as Haraway insisted, remains political—and human—all the way down.
Annotated Bibliography
Primary Sources - Donna Haraway
Haraway, Donna. "A Cyborg Manifesto: Science, Technology, and Socialist-Feminism in the Late Twentieth Century." Reprinted in Simians, Cyborgs and Women: The Reinvention of Nature. New York: Routledge, 1991. The foundational text introducing cyborg theory as a framework for understanding human-machine relationships beyond traditional dualisms. Essential for understanding boundary dissolution concepts.
Haraway, Donna. "Situated Knowledges: The Science Question in Feminism and the Privilege of Partial Perspective." Feminist Studies 14, no. 3 (1988): 575-599. Develops the epistemological framework of situated knowledges as an alternative to both positivist objectivity and relativist constructivism. Crucial for understanding AI bias and training data issues.
Haraway, Donna. Staying with the Trouble Durham: Duke University Press, 2016. Later work developing multispecies thinking and response-ability concepts. Essential for understanding environmental and ecological approaches to AI development.
Contemporary Applications to AI
Trächtler, Jasmin. "The world as witty agent—Donna Haraway on the object of knowledge." Frontiers in Psychology 15 (2024): 1389575. Comprehensive 2024 analysis applying Haraway's situated knowledges framework to contemporary AI systems and scientific objectivity debates.
Draude, C., G. Klumbyte, P. Lücking, and P. Treusch. "Situated algorithms: a sociotechnical systemic approach to bias." Online Information Review 44, no. 2 (2020): 325-342. Influential study directly applying situated knowledges framework to algorithmic bias, proposing situated algorithms as alternative to universal AI systems.
Lewis, Jason Edward, Noelani Arista, Archer Pechawis, and Suzanne Kite. "Making Kin with the Machines." In Atlas of Anomalous AI, edited by Ben Vickers and K Allado-McDowell, 2020. Groundbreaking work extending companion species concepts to AI relationships through indigenous epistemological frameworks.
Toupin, Sophie. "Shaping feminist artificial intelligence." New Media & Society (2024). Comprehensive survey of feminist AI development, documenting applications of Haraway's theoretical frameworks to contemporary AI research.
Technical and Policy Sources
European Union. "Artificial Intelligence Act." Official Journal of the European Union, 2025. Primary regulatory framework demonstrating situated approaches to AI governance.
UNESCO. "Recommendation on the Ethics of Artificial Intelligence." 2021. Global policy framework establishing ethical principles for AI development.
World Economic Forum. "Future of Jobs Report 2025." 2025. Analysis of AI's impact on human labor and collaboration.
Stanford University Human-Centered AI Institute. "AI Index Report 2025." 2025. Comprehensive data on AI development trends, capabilities, and social impacts.
Environmental and Social Impact
Climate Change AI. "Tackling Climate Change with Machine Learning." 2022. Analysis of AI applications for environmental monitoring and climate change mitigation.
International Energy Agency. "Data Centers and Energy From Global Perspectives." 2023. Assessment of AI infrastructure's environmental impact and energy consumption.
UNESCO. "I'd blush if I could: closing gender divides in digital skills through education."