WASHINGTON, April 13, 2026 – Criticism of artificial intelligence has turned violent in San Francisco after two attacks on OpenAI CEO **Sam Altman** ’s home within the last four days. At about 2:56 a.m. on Sunday, San Francisco police responded to possible shots fired in Russian Hill, as first reported by The San Francisco Standard. A person in the passenger seat fired a round from the car window, according to a police report. While the car fled, a camera captured the license plate, allowing SFPD to detain 25-year-old San Franciscan **Amanda Tom** and 23-year-old **Muhamad Tarik Hussein**. According to a police report, the police seized three firearms from the suspects as they were booked for negligent discharge. This was the second violent incident on Altman’s house, with the first happening just two days earlier on Friday. A 20-year-old Texas resident **Daniel Alejandro Moreno-Gama** threw a Molotov cocktail — a bottle containing a flaming rag and a flammable liquid — at the property’s gate just two days earlier. Moreno-Gama, a critic of AI, was arrested later that day and charged for suspicion of attempted murder, arson, possession or manufacture of an incendiary device, and more. Learn more about the Broadband Community... Start Your Broadband Journey Here ### This post is for subscribers only Become a member to get access to all content Subscribe now

In January, the winners of the Berggruen Prize were announced. For those who are not familiar with this prize, it is considered the most important philosophy essay competition. Anil Seth, a famous neuroscientist, won the English-language section with an essay entitled “The Mythology of Conscious AI.” When the essay was published in the magazine Noema, some of my readers asked me about my views on it. That is a fair ask, given that debates on AI consciousness are increasingly splitting the research community into proponents and opponents of the possibility of such a creation, while setting aside the ethical questions that would follow if it were possible. This is a long text. In this first post, I would like to present the beginning of Anil Seth’s essay. The essay opens by explaining that humans have long sought ways to create human-like beings. Anil Seth adds that the creation of conscious beings has also been an important topic in human cultural history. He writes, for instance: > “The cultural history of synthetic consciousness is both long and mostly unhappy. From Yossele the Golem, to Mary Shelley’s “Frankenstein,” HAL 9000 in “2001: A Space Odyssey,” Ava in “Ex Machina,” and Klara in “Klara and The Sun,” the dream of creating artificial bodies and synthetic minds that both think and _feel_ rarely ends well — at least, not for the humans involved. One thing we learn from these stories: If artificial intelligence is on a path toward real consciousness, or even toward systems that persuasively seem to be conscious, there’s plenty at stake — and not just disruption in job markets.” Seth reminds us — _was it necessary?_ — that voices stating that conscious AI already exists, while others consider such a statement as misleading, or false, or as pertaining to the SF register. Still, he concludes: “For many leading experts in AI and neuroscience, the emergence of machine consciousness is a question of _when,_ not _if_.” What is clear here is that, even though many leading experts agree, that does not necessarily mean they are right in their judgment. That is why Seth feels important to add here that: > “How we think about the prospects for conscious AI matters. It matters for the AI systems themselves, since — if they are conscious, whether now or in the future — with consciousness comes moral status, the potential for suffering and, perhaps, rights.” He also added that it matters to us because it may change how we consider, use, or even interact with our AI tools. It matters also because it may change the way we understand “our own human nature and the nature of the conscious experiences that make our lives worth living.” There’s nothing new here, but it is a good, general public-oriented introduction. The body of the text begins with a section titled “The Temptations Of Conscious AI,” which addresses the question of whether AI could be conscious and explains why it is an important question. Then, it presents four arguments related to “Consciousness & Computation,” the field of expertise of Anil Seth. Next is a section oriented more toward ethical questions titled “What (Not) To Do,” followed by a conclusion on “Soul Machine.” From now on, I will restitute the argumentation he presented following his text. To facilitate comparison with Anil Seth’s own text, I will use his section titles for the present discussion. ### This post is for subscribers only Become a member to get access to all content Subscribe now

Five years after its publication in 2021, _The Age of AI: And Our Human Future_ reads less like a “future-oriented” book and more like an account of what has since become reality. In their bestselling book on artificial intelligence, Henry Kissinger, Eric Schmidt, and Daniel Huttenlocher recognized that the decisive change would not be a gadget or a single breakthrough, but rather a transformation in how our world is organized. According to the authors, this transformation encompasses several questions, such as: > “What do Al-enabled innovations in health, biology, space, and quantum physics look like? What do Al-enabled “best friends” look like, especially to children What does Al-enabled war look like? Does Al perceive aspects of reality humans do not? When AI participates in assessing and shaping human action, how will humans change? What, then, will it mean to be human?” (p. 5) These questions are fundamentally philosophical and political. They explore what constitutes knowledge, agency, and responsibility — even “reality” — when AI enters domains such as medicine, warfare, and scientific discovery. These questions, in turn, raise ethical, governance, and social issues. Such questions also address epistemological and metaphysical concerns, such as whether AI can detect patterns or dimensions of the world that humans cannot and what that would mean for human self-understanding. They converge on a central anthropological question: If AI helps assess and direct human actions, how will autonomy or identity evolve, and what will remain distinctive about being human? ### Five Years On The book is structured as a series of expanding circles. The seven chapters progress from an overview of “Where We Are” to the ethical and institutional implications of “AI and the Future.” This structure reflects the authors’ idea that AI will change how humans experience reality and perform more tasks, affecting various aspects of life and society. The authors consider this change comparable to epochal revolutions, such as the Enlightenment’s celebration of reason. They argue that AI is not “an industry” or “a single product,” but a technology that will permeate research, education, logistics, defense, culture, and everyday decision-making because it can “learn, evolve, and surprise” (p. 4). They present opening examples, such as deep learning–based antibiotic discovery at the Massachusetts Institute of Technology (MIT), as metaphors for a new kind of cognition operating alongside our own. Regarding MIT’s use of AI in their research on new antibiotics, they emphasize that > […] by training a software program to identify structural patterns in molecules that have proved effective in fighting bacteria, the identification process was made more efficient and inexpensive. The program did not need to understand why the molecules worked — indeed, in some cases, no one knows why some of the molecules worked. Nonetheless, the AI could scan the library of candidates to identify one that would perform a desired albeit still undiscovered function […]. (p. 10) And they add: > The AI that MIT researchers trained did not simply recapitulate conclusions derived from the previously observed qualities of the molecules. Rather, it detected new molecular qualities — relationships between aspects of their structure and their antibiotic capacity that humans had neither perceived nor defined. (p. 11) The discovery of this new antibiotic, called Halicin, suggests that machines can devise strategies humans neither programmed nor conceived. This also forces us to ask what the machine has identified in a bounded domain that we have not, and what this or other AI might figure out that we have not yet (and possibly will not ever be able to without AI assistance). ### Knowledge After Reason The book’s most interesting contribution is its assertion that AI is fundamentally epistemological. It questions what constitutes knowledge and the process of transforming information into knowledge. According to the authors, the internet has already inundated the mind with decontextualized fragments, and AI systems — especially those embedded in search, recommendation, and personalization — complete this transformation by becoming permanent companions in the perception and processing of information. The authors’ philosophical framework is unusually explicit for a bestselling book on technology. Among other references, Descartes, Kant’s “thing-in-itself,” Wittgenstein’s “family resemblances,” and twentieth-century physics — which destabilized naive objectivity — all serve as lenses for understanding AI’s alternative access to reality. The authors’ recurring question, whether AI approaches the same reality from a different standpoint or reveals partially overlapping realities, gives the book its haunting tone. Henry Kissinger’s influence is most evident at the intersection of AI and strategy, particularly in deterrence and escalation. There is a growing concern that militaries may adopt tactics influenced by pattern recognition that surpass human calculation. The book emphasizes that traditional concepts of deterrence and even the laws of war may deteriorate or require fundamental adaptation if autonomous or semi-autonomous systems begin to select targets or execute engagements. The book also ties cyber operations to unpredictability. As AI is grafted onto cyber weapons, their discriminating potential and capacity for cascading damage coexist, blurring the distinction between conventional and exceptional force. ### The Debate They Missed ### This post is for subscribers only Become a member to get access to all content Subscribe now

#### The Worldview Behind Altman’s Idea of a “Better Life” In the January 7, 2025, episode of _ReThinking with Adam Grant_ titled “Sam Altman on the Future of AI and Humanity,” Adam Grant concludes with a deceptively simple question for the OpenAI CEO: > “With a child on the way, as a soon to be father, what kind of world are you hoping to see for the next generation?” In his interview, he offers more than just a personal response. He offers a concise worldview about what constitutes a “better life” in the age of AI, something that is worth looking at in detail. In this essay, I argue that Altman’s response reveals what we can call the “faith of the technologist”: the belief that scientific and technological progress, especially AI-driven progress, is the primary driver of prosperity, fulfillment, and human advancement. To illustrate this idea, I draw on Isabelle Stengers’s concept in _Cosmopolitics_ (2010) that modern scientific projects depend on specific forms of “faith” that determine what constitutes reality and what is excluded. I will approach this in three steps. First, I will unpack what “a better life” means in Altman’s framework. Second, I will use Stengers to demonstrate how this concept functions as a form of faith. Third, I will explore what this cosmology tends to omit. ### This post is for subscribers only Become a member to get access to all content Subscribe now

In 2024, Peter Königs published a widely discussed paper titled “In Defense of Surveillance Capitalism.” A year later, in November 2025, he released another influential work addressing what he called the “negativity crisis” in AI ethics. He argues that philosophical debates on artificial intelligence almost always focus on problems and risks. By contrast, positive or constructive perspectives on AI appear rarely. Königs emphasizes that, although AI holds potential to improve human life, the ways in which it might do so are often overlooked by ethicists. Thus, his main claim is that ethical discourse on AI suffers from a pervasive negativity that distorts how we understand the technology’s impact. Instead of defending AI against its critics, Königs examines why negativity dominates by analyzing the field from a philosophy of science standpoint. He attributes this one-sided focus to three interrelated factors: 1. AI ethics engages emerging technologies rather than timeless questions, positioning philosophers as commentators reacting to innovation. 2. There exists an implicit academic norm: papers that explore ethical benefits rather than harms are often seen as insufficiently novel or prescriptive. 3. Institutional pressures — especially the need to publish and obtain funding — encourage scholars to emphasize risks, because addressing concerns signals relevance, while optimistic accounts may seem less serious. ### This post is for subscribers only Become a member to get access to all content Subscribe now

In debates about influencer marketing, the focus is usually on mega-influencers and celebrities with millions of followers. However, the paper “Less is more: Engagement with the content of social media influencers” by Jesse Pieter van der Harst and Spyros Angelopoulos challenges this narrative. It argues that, when it comes to genuine engagement, size is not everything. Micro‑influencers receive more favorites per follower than larger influencers, while larger accounts generate more sharing and total engagement because of their reach. ### Rethinking Influence The paper begins by situating influencers within the broader context of celebrity endorsements, from radio and television to modern social media platforms. While the traditional account suggests that larger audiences are always better, the authors demonstrate that this idea is too simplistic when it comes to social media. They distinguish four tiers of influencers on X: micro (5,000–50,000 followers), meso (50,000–100,000 followers), macro (100,000–1,000,000 followers), and mega (>1,000,000 followers). Then, they ask, “**Who generates more engagement per follower?** ” To answer this question, the authors compiled a large dataset of self-identified influencers or brand ambassadors on X. They collected the most recent English-language posts from each account’s timeline from early 2021. They measured engagement in two basic ways: favorites (likes) and retweets (shares), in both their raw form and normalized by follower count. This allowed the authors to distinguish the intensity of engagement per audience member from volume. ### This post is for subscribers only Become a member to get access to all content Subscribe now

In many contemporary debates, “surveillance capitalism” often serves as a convenient culprit, being blamed for political decline, mental health crises, and erosion of privacy. However, in his paper “In Defense of Surveillance Capitalism,” Peter Königs argues that this portrayal is empirically fragile and normatively one-sided. As he puts it: > “Many critical discussions of surveillance capitalism suffer from two defects. They often exaggerate the negative aspects of surveillance capitalism, and they fail to acknowledge its positive aspects.” ### A Critique of a Critique The paper begins by clarifying the meaning of “surveillance capitalism.” It focuses on companies such as Google, Meta, YouTube, and X, which offer free digital services funded by collecting data and selling advertising. Königs notes that the term “surveillance capitalism” is misleading because if surveillance is defined as monitoring to enforce norms or detect wrongdoing, then most of what these firms do — profiling users to show them relevant ads — does not qualify as surveillance. He explains: > “We usually do not speak of surveillance unless the monitoring or data-collecting serves the purpose of verifying that people comply with certain normative expectations, such as laws, moral norms, orders, etc.” Similarly, labeling the economy as a new “mutation” of capitalism dominated by these practices exaggerates their influence, considering data-driven advertising remains a relatively small sector. Königs therefore treats “surveillance capitalism” primarily as a label for a specific business model (data-funded, advertising-supported services), not as a fundamentally new economic order. ### This post is for subscribers only Become a member to get access to all content Subscribe now

In recent years, the question of whether we can trust artificial intelligence (AI) has moved from philosophical seminars to boardrooms and legislative chambers. Trust is often described as the “foundation of interpersonal interactions” and a “fundamental component of a functional society.” As we integrate AI into our lives, from chatbots to autonomous decision systems, we ask: Is this system trustworthy? In a new paper published in _AI and Ethics_, Jonathan Tallant and I argue that we are asking the wrong question. By focusing on “trustworthiness” — whether an AI system is worthy of trust — we skip a crucial step: “trustability.” Absent from the philosophical literature, this concept defines whether an entity is even the kind of thing that can be trusted. We argue that treating AI as a candidate for trust is not just risky, but also a category error with profound ethical consequences. ### The Trustability Gap The distinction between “trustability” and “trustworthiness” is the paper’s central conceptual contribution. * **Trustworthiness** : is a normative evaluation: Is this agent honest, competent, and reliable? It assumes the agent is capable of entering a trust relationship. * **Trustability is a categorical threshold** : Is this entity capable of holding the moral and structural qualities required for trust to exist at all? Drawing on Paul Faulkner’s “grammar of trust” and recent work by Massaguer Gómez on human-robot interaction, we argue that current AI systems fail this first test. They elicit feelings of trust — through natural language, polite interfaces, and confident answers — without possessing the internal capacity to be _trustable_. They are mimics of trustworthiness, operating in a vacuum of accountability. ### From Misplaced Trust to Structural Incoherence When we trust an untrustable entity, we are not merely making a bad bet. We are engaging in a “structurally incoherent” attitude. Our paper posits that trust is not just a prediction of behavior (which would be mere “reliance”) but a normatively charged stance. We feel “betrayed” when a trusted person fails us. When a machine fails, we may be disappointed or harmed, but we cannot be “betrayed” in the moral sense because the machine was never a moral agent to begin with. This confusion leads to problematic design and governance outcomes. If we treat AI as a “trust partner,” we might try to solve failure modes by “building trust” (e.g., making the AI sound more empathetic) rather than “ensuring reliability.” In our paper, we warn that many AI systems today are designed precisely to obscure this distinction, encouraging users to project human-like accountability onto software that cannot reciprocate it. ### Reliance with Accountability If we cannot “trust” AI, what should we do? We propose shifting our framework from “trust” to “reliance with accountability.” * **Reliance is instrumental** : We rely on a car to start or a bridge to hold weight. If it fails, we fix the engineering. No moral relationship exists between us and the tool. * **Trust is relational and requires agenc** y: It involves a relationship between moral agents — beings capable of understanding obligations, making choices, and being held accountable for their actions. When we trust someone, we implicitly assume they can comprehend what we’re trusting them with and can choose to honor or betray that trust. By categorizing AI strictly under “reliance,” we clarify the ethical landscape. The key point is not about our vulnerability to machines — we are vulnerable to bridges and cars too — but about the absence of moral agency in AI. A machine cannot understand betrayal because it cannot make a moral choice. It simply fails or succeeds according to its programming. The burden of responsibility therefore shifts from the machine (which cannot bear it) to the institutions, developers, and policymakers who deployed it. They are the moral agents. Our paper examines how this aligns with debates on trust in governments and institutions. While AI itself cannot be a trustee, the systems surrounding it can be held to standards that make reliance rational. ### Conclusion: The Conditions for Future Trust The paper concludes with an exploration of whether AI could ever become trustable. We outline the technical, moral, and political prerequisites necessary to reach this threshold. Simply making models more accurate is not enough. For an AI to be genuinely trustable, and possibly trustworthy, it would require a status that integrates it into our moral expectations in a way that does not currently exist: by a recognition of dependence and accountability. Moreover, trustworthy AI would also depend on the quality of AI integration in social, political, and organizational landscapes—on whether the systems are embedded in structures of genuine oversight, accountability, and redress. Until then, trust in AI is a misnomer. Policymakers and designers must diagnose where trust is structurally impossible and replace it with rigorous, verifiable reliability rather than trying to humanize our machines. Instead of trying to forge a relationship with our tools, we must take responsibility for how we use them.

A clothing purchase, a consulting report line, or an AI prompt seems insignificant alone. Yet each participates in networks where minor choices cascade through supply chains, institutions, and people’s lives — creating a butterfly effect. This article explores cascades in three domains — fashion, consulting, and the meaning of work — to show how governance and behaviors must adapt when minor decisions have the power to impact entire systems. ## Fashionable Ethics Under Strain Current debates about fashion illustrate the matter in detail. A special issue of the _Journal of Business Ethics_ sheds light on the contrasting nature of the fashion industry: enchanting yet exploitative. Problems include labor abuses, environmental damage, cultural appropriation, and discrimination. Manufacturing is concentrated in the Global South, where protections are weak. Exploitation, slavery, environmental damage, and waste colonialism are not anomalies — they are systemic. Cost-cutting and speed directly result in longer hours, dangerous conditions, and new environmental harms. Brands must extend their responsibility to include not only direct employees, but also suppliers and vulnerable workers, especially during crises like the pandemic. Firms engage in “techwashing” — exaggerating sustainability claims. Meaningful change requires NGOs, consultants, agencies, firms, and regulators to work together on information, accountability, and risk distribution. Marketing exacerbates problems. Social media encourages impulse buying, campaigns enforce narrow beauty standards, and provocative ads reinforce harmful stereotypes and gender-based violence — all for short-term sales gains. ## Deloitte and Ai’s Reliability Problem The same problems — issues that cascade into trust failures — exist beyond fashion. For example, in 2025, Deloitte returned fees after a legal scholar discovered fabricated quotes and fake sources in its Australia welfare report. The firm had used OpenAI tools to draft sections without proper verification, though the main recommendations survived. The real problem is that organizational procedures failed to detect the fabrications. What seemed like a shortcut became a test case for institutional integrity. Reports influence decisions even when they are only partially read because they are trusted to verify facts and methods. Unverified AI-generated statements undermine this institutional authority. This raises questions about internal controls, client oversight, and systems of professional accountability. In this case, the butterfly effect operates through expectations rather than supply chains. A few unchecked citations can undermine confidence in entire levels of review, even if the primary analysis remains unchanged. Structural pressures suggest that more incidents are likely. Organizations are strongly incentivized to quickly integrate AI to remain efficient and competitive. However, developing documentation, testing protocols, and human review processes for AI tools is slow and resource-intensive. This does not mean that generative AI is unusable. Rather, it means that the reliability of generative AI depends on the wider socio-technical system in which the models are embedded. In the Deloitte case, the notable failure was not that hallucinations occurred, as this is expected, but rather that they were not identified by processes that should have treated the AI outputs with appropriate skepticism. ## McKinsey’s Superagency Vision McKinsey’s 2025 report on “Superagency” reveals a gap: companies claim AI ambition but lack maturity in execution. While a substantial majority of companies intend to increase their investment in AI, only a small percentage of leaders consider their organizations to be mature in terms of AI integration into workflows and its impact on business outcomes. Interestingly, employees report higher levels of AI use than leaders estimate, and many expect their use of generative tools to expand quickly. This suggests that bottom-up experimentation is outpacing formal strategy, and employees are leading, not resisting, change. The report portrays AI as a liberating technology. Like the internet, AI has the potential to democratize knowledge and automate routine work if it is deployed safely and equitably. Although the framework rightly emphasizes governance over algorithms, it falls short in practice. Self-reported data introduces bias, training-focused solutions overlook accountability gaps, and leaders prioritize operational benchmarks and performance metrics over ethical/compliance benchmarks when measuring AI systems. The document’s origins also shape its outlook. As a consulting document linked to a book promoting the idea of superagency, the report is designed to encourage investment and frame AI as a strategic opportunity that firms cannot ignore. This does not invalidate the findings, but it highlights executive action and competitive advantage over everyday maintenance work, sector-specific regulations, and social risks that make AI adoption ethical or problematic in practice. ## Existential Unemployment and Meaningful Work These cascades raise a deeper question. When AI outperforms humans in meaningful work, what remains valuable about being human? O’Brien’s article shifts the focus from productivity to meaning. If AI outperforms humans in research, philosophy, and art, three threats emerge: altered incentives that discourage people from entering demanding fields, reduced skills through outsourcing cognitive work, and the transformation of meaningful work into trivial activities, even when humans continue to perform them. A single breakthrough by a machine can alter how people perceive the potential of human contributions, influencing how younger generations evaluate the value of embarking on long and arduous journeys in research or the arts. O’Brien distinguishes between automating routine tasks and automating meaningful work — research, art, creation — which hold objective and personal value. Proposed solutions range from embracing games and expanding care work to restricting AI or enhancing human capacity. Each solution has its limitations. Care work is meaningful, but it cannot replace intellectual creation. Appreciation matters, but it differs from achievement. Deskilling erodes passive engagement as well. Restricting AI protects meaning, but it also delays medical and environmental solutions. Enhancement technologies may lag behind automation, creating interim periods of “temporary obsolescence” that reshape values. O’Brien acknowledges the moral and economic costs of delaying AI, such as foregoing medical and environmental solutions, but he argues that these costs may be worth bearing. In the end, he concludes that the threat to meaning provides us with a prima facie reason to slow down AI development and opt for patience over rapid automation. ## Conclusion: Ethics Across Systems Today’s crucial ethical questions are not about dramatic failures — they are about how small, routine choices accumulate. A quick purchase, a rushed edit, a metrics choice, a AI shortcut — each seems trivial. Yet in interconnected systems, they cascade into labor abuses, governance failures, and cultural shifts no one intended. Ethical governance requires noticing and designing for cause-and-effect chains. This requires treating small choices as consequential and protecting human judgment from technological replacement.

For many fashion brands, ethical initiatives offer the promise of enhancing brand reputation, satisfying consumer demand, and meeting regulatory requirements. However, empirical research suggests that the success of an ethical transformation depends on more than just governance commitment, but also on the interconnectedness of systems, stakeholders, and ethical frameworks across the entire value chain. “Fashionable Ethics: Exploring Ethical Perspectives in the Production, Marketing, and Consumption of Fashion,” a special issue of the _Journal of Business Ethics_ , provides a rich framework that illustrates this point. The authors introduce the challenge of implementing ethical practices as follows: > “Fashion is simultaneously enthralling yet exploitative, replete with a multitude of ethical issues along the entire value chain from production and marketing to consumption, incorporating labor exploitation, animal cruelty, environmental pollution, consumerism, cultural appropriation, objectification, under-representation, and discrimination.” ### Why This Special Issue Matters The editors note that “research on ethical issues in fashion is growing but is fragmented across diverse domains,” with “only a peripheral focus on ethics and limited application of ethical theories or frameworks to fashion’s ethical dilemmas.” This special issue deliberately “brings together diverse domains and unpacks salient ethical issues using the lens of ethical theories and frameworks” to better balance “social justice with environmental responsibility, addressing consumerism and new forms of greenwashing, cultural appropriation, objectification, under-representation, and discrimination.” Across nine papers, the authors mobilize deontology, virtue ethics, Confucian virtue ethics, contractualism, and related perspectives to analyze how real actors — employees, entrepreneurs, consumers, and suppliers — navigate ethical tensions in production, marketing, and consumption. ### Production: Ethics Under Pressure On the production side, the special issue emphasizes that “the complex nature of global fashion supply chains presents significant ethical challenges, particularly in terms of traceability, transparency, and the multi-tiered structure of these networks,” with most production “concentrated in the Global South where cheap labor is abundant” and “institutional voids” undermine worker protection. Both developed and developing countries have systemic issues such as exploitative working conditions, “modern slavery,” environmental damage, groundwater depletion, and waste colonialism — which is a new form of colonialism in which waste and pollution are used to dominate a group of people in their homeland. These challenges are presented as systemic outcomes of current business models rather than isolated abuses. The special issue emphasizes that brands cannot limit their responsibility to their employees. For fashion brands to embrace ethical practices, they must prioritize the safety, security, and livelihoods of their suppliers and workforce — the most vulnerable members of the fashion supply chain. It also highlights how crises expose structural injustices. During the pandemic, widespread cancellations and payment defaults threatened established supply chains and imposed severe economic and social pressures on vulnerable workers. Newer phenomena, such as “techwashing,” exemplify how brands misrepresent or exaggerate technological advancements to appear more innovative or sustainable than they truly are. Meanwhile, corruption and opaque procurement practices further erode trust and fairness. ### This post is for subscribers only Become a member to get access to all content Subscribe now

In many organizations, the promise of introducing advanced technologies centers on automation, cost reduction, and speed. However, empirical research suggests that a key factor in the success of new technologies is not only the capability of the machine but also how humans, systems, and technologies interconnect. A study by Brynjolfsson, Li, and Raymond titled “Generative AI at Work,” published in _The Quarterly Journal of Economics_ , provides a rich case study that illustrates this point. The authors introduce the question of implementing AI technologies as follows: > "The emergence of generative artificial intelligence (AI) has attracted significant attention for its potential economic impact. Although various generative AI tools have performed well in laboratory settings, questions remain about their effectiveness in the real world, where they may encounter unfamiliar problems, face organizational resistance, or provide misleading information (Peng et al. 2023a; Roose 2023)." ### This post is for subscribers only Become a member to get access to all content Subscribe now

In January 2025, McKinsey & Company released _Superagency in the Workplace: Empowering People to Unlock AI’s Full Potential_, a report that has already influenced how business leaders discuss artificial intelligence. The report captures a paradox at the heart of today’s workplace. Nearly every company is investing in AI, yet very few are realizing its potential. As the report puts it: > “Therein lies the challenge: the long-term potential of AI is great, but the short-term returns are unclear. Over the next three years, 92 percent of companies plan to increase their AI investments. But while nearly all companies are investing in AI, only 1 percent of leaders call their companies “mature” on the deployment spectrum, meaning that AI is fully integrated into workflows and drives substantial business outcomes. **The big question is how business leaders can deploy capital and steer their organizations closer to AI maturity.”**(p. 2, my emphasis) They also bring up this issue when they ask: “How can companies harness AI to amplify human agency and unlock new levels of creativity and productivity in the workplace?” (p. 2). The report’s main argument is that the issue is not a technological one, but one of governance. In this piece, I will provide an overview of the report and its limitations. One thing should be noted first. The report does not aim to question AI or its uses. Rather, the report aims to present the opportunities AI may offer businesses and the challenges of implementation. This is a consulting report. This explains why the report includes paragraphs like the following: > “Imagine a world where machines not only perform physical labor but also think, learn, and make autonomous decisions. This world includes humans in the loop, bringing people and machines together in a state of superagency that increases personal productivity and creativity (see sidebar “AI superagency”). This is the transformative potential of AI, a technology with a potential impact poised to surpass even the biggest innovations of the past, from the printing press to the automobile. AI does not just automate tasks but goes further by automating cognitive functions. Unlike any invention before, AI-powered software can adapt, plan, guide — and even make — decisions. That’s why AI can be a catalyst for unprecedented economic growth and societal change in virtually every aspect of life. **It will reshape our interaction with technology and with one another.”** (p. 6, my emphasis) Keeping in mind that the authors aim to present AI opportunities in a positive light and largely set aside the potential social, political, and cognitive difficulties these technologies and their use may cause, let’s examine the report itself. ### What the Report Says and Shows The survey was conducted in October and November of 2024. The report is based on extensive empirical research, including a survey of 3,613 employees (“managers and independent contributors”) and 238 C-suite leaders (top-ranking executives), primarily (81%) in the United States and five other countries: Australia, India, New Zealand, Singapore, and the United Kingdom (19%). Although the sample includes respondents outside of the US, the findings discussed in the report “pertain solely to US workplaces.” The employees surveyed represented many functional areas, “including business development, finance, marketing, product management, sales, and technology.” As indicated in the introduction to this article, the data shows that enthusiasm is outpacing strategy overall. 92% of companies plan to increase AI investment over the next three years, yet only 1% of leaders call their companies “mature,” “meaning that AI is fully integrated into workflows” and driving “substantial business outcomes.” In other words, most companies were not yet “mature” — AI had not yet been fully integrated into workflows and was not yet driving substantial business outcomes. The gap between ambition and execution is huge. However, according to the same report, employees are advancing more quickly than leaders realize. While leaders estimate that only about four percent of employees use generative AI for at least a third of their work, employees report a rate three times greater. Furthermore, 47% of employees — versus 20% of leaders — expect to use generative AI at that level within a year. The authors of the report came to a simple conclusion based on this situation: employees are not resisting change, but rather, leading it. Employees aged 35–44 emerge as the most confident adopters, often acting as informal “AI help desks” within their teams. On top of this, most want formal training and sanctioned opportunities to experiment. The authors argue that businesses must think bigger and act faster. They should set concrete, outcome-focused goals. They should also invest in training, establish governance early on, and include non-technical employees in the ideation process. According to the report, the end state is “superagency,” a condition in which technology amplifies human creativity and decision-making rather than replacing them. On this concept, they write: > “Superagency, a term coined by Hoffman, describes a state where individuals, empowered by AI, supercharge their creativity, productivity, and positive impact. Even those not directly engaging with AI can benefit from its broader effects on knowledge, efficiency, and innovation. AI is the latest in a series of transformative supertools, including the steam engine, internet, and smartphone, that have reshaped our world by amplifying human capabilities. Like its predecessors, **AI can democratize access to knowledge and automate tasks, assuming humans can develop and deploy it safely and equitably**.” (p. 6, my emphasis) The idea of equitably deploying AI is more of a slogan than a practical concept. A consulting firm like McKinsey & Company does not aim to provide equitable capabilities to all companies; rather, it aims to provide competitive advantages to its clients. The rhetoric serves one goal: advising clients and potential clients not to miss the technological shift that AI will bring about. ### What It Implies for Today’s Businesses For contemporary businesses, the implications of AI are twofold. First, the productivity potential is real. Independent field studies corroborate the report’s optimism. A large experiment on customer-support agents showed that a GPT-based assistant increased productivity by around fifteen percent, with the biggest improvements among novices (Erik Brynjolfsson _et al._ 2025). Second, scaling AI is not a data-science problem but an organizational one. As the authors of the report put it: > “Achieving AI superagency in the workplace is not simply about mastering technology. It is every bit as much about supporting people, creating processes, and managing governance.” (p. 10) The report also shows that, although many employees are concerned about cybersecurity, inaccuracy, and privacy, they trust their employers more than other institutions to use AI responsibly. In this case, higher employee trust reduces hesitation and supports the adoption of AI. Beyond that, the authors call for monitoring AI for fairness, safety, and explainability and note that few leaders who benchmark AI consider it important. They write: > “One powerful control mechanism is respected third-party benchmarking that can increase AI safety and trust. (…) While benchmarks have significant potential to build trust, our survey shows that only 39 percent of C-suite leaders use them to evaluate their AI systems. Furthermore, when leaders do use benchmarks, they opt to measure operational metrics (for example, scalability, reliability, robustness, and cost efficiency) and performance-related metrics (including accuracy, precision, F1 score, latency, and throughput). These benchmarking efforts tend to be less focused on ethical and compliance concerns: Only 17 percent of C-suite leaders who benchmark say it’s most important to measure fairness, bias, transparency, privacy, and regulatory issues (…).” (p. 24) Taken together, these findings demonstrate that leaders tend to focus their third-party benchmarking efforts on operational and performance metrics. This limits their ability to address the ethical and compliance issues associated with AI use. Therefore, current evaluations must address concerns such as fairness, bias, transparency, privacy, and regulatory issues. Recognizing the importance of these dimensions requires developing the necessary capabilities. Leaders need training and skills to ensure the implementation of AI technology is of high quality. Again, this comes down to a governance issue. ### Where the Report Falls Short McKinsey’s framework does have its limitations. Since its data comes from self-reporting, adoption rates and impact estimates are subject to optimism bias. Employees may overstate the percentage of their work that is genuinely mediated by AI, while leaders may understate the amount of informal use. The resulting perception gap may be significant. While justified, the emphasis on training risks oversimplifying what actually enables responsible performance. Training only teaches users how to prompt. However, it does not redesign workflows to ensure that AI outputs are reviewable and that errors are correctable. Meanwhile, fairness is mainly treated as a technical issue, but fairness also depends on whether AI-assisted decisions feel fair, are explainable, and can be appealed (Jabagi _et al._ 2025). The report pays little attention to the everyday maintenance work that makes AI viable. Behind each successful use case are hours of human correction and quality control, as well as dialogue with clients or colleagues to interpret the system’s output. It is essential to recognize this labor if “superagency” is to describe empowerment rather than an additional burden. This work is also instrumental in the general acceptability of AI. One final limitation concerns the origin and context of the report itself. _Superagency in the Workplace_ is a McKinsey & Company report written by practitioners. It was explicitly “prompted by” the book _Superagency: What Could Possibly Go Right with Our AI Future_ (Authors Equity, January 2025). This context helps explain the report’s rather optimistic tone and focus on leadership strategies, such as its call for respected third-party benchmarking. While this does not invalidate the findings, it can shift the focus toward executive action and operational metrics, diverting attention from more complex labor processes and sector-specific regulatory barriers. Readers should keep this in mind when interpreting the data and recommended actions. ### The Bottom Line This report is best read as both diagnosis and provocation. Its data reveal momentum — employees are experimenting faster than leaders assume — and its optimism rests on a plausible structure: ambition combined with effective governance. Yet “superagency” should not be mistaken for a technological state of grace. True agency lies in the institutionalization of good habits: clarifying responsibilities, making systems inspectable, and ensuring that those who rely on AI remain able to question and correct it. For today’s organizations, the question is not whether AI will augment human capacity but whether they can build the conditions that make such augmentation ethical as well as sustainable. The challenge is cultural, organizational, and ethical — how to ensure that the future of work amplifies judgment rather than replacing it with unexamined certainty. The authors make the same case by looking back at the last major shift, then urging leaders to act now. > “AI could drive enormous positive and disruptive change. This transformation will take some time, but leaders must not be dissuaded. Instead, they must advance boldly today to avoid becoming uncompetitive tomorrow. The history of major economic and technological shifts shows that such moments can define the rise and fall of companies. Over 40 years ago, the internet was born. Since then, companies including Alphabet, Amazon, Apple, Meta, and Microsoft have attained trillion-dollar market capitalizations. Even more profoundly, the internet changed the anatomy of work and access to information. AI now is like the internet many years ago: **The risk for business leaders is not thinking too big, but rather too small.”** (p. 2)

Leadership experts often encourage managers to empower their subordinates by delegating important tasks, granting autonomy, and, above all, trusting their team. The logic seems intuitive: when people feel trusted, they perform better and become more engaged. However, a new study published in the _Journal of Business Ethics_ complicates this idea by showing that aligning trust with employees’ desires can inadvertently promote unethical loyalty in the absence of ethical guidance. ### Trusting the Right Amount Drawing on social exchange theory and moral disengagement theory, the authors hypothesize that when supervisors provide employees with the level of trust they desire — especially at high levels — employees may feel compelled to reciprocate this trust, even by engaging in unethical acts. The alignment between desired and received trust, referred to as “trust congruence,” is expected to strengthen the social exchange relationship between supervisor and subordinate. Under conditions of weak ethical leadership, this strengthened exchange may encourage moral disengagement, causing employees to justify behavior such as lying or hiding mistakes for the benefit of their supervisor. The authors term such conduct _unethical pro-supervisor behavior._ This perspective challenges the prevailing notion that the “right amount of trust” always results in positive outcomes. While previous research shows that perceived trust enhances performance and commitment, other studies on moral licensing suggest that feeling valued can sometimes prompt individuals to take ethical shortcuts. The authors therefore ask: > Does feeling trusted encourage loyalty, even if it means sacrificing integrity? ### Four Studies Across Three Countries The researchers investigated whether “trust congruence” — the alignment between the trust employees desire from their supervisor and the trust they actually receive — could unintentionally encourage _unethical pro-supervisor behavior_. Across four studies — one experiment in the U.S. and three time-lagged field studies in France, the U.K., and the U.S. — the authors tested and confirmed their model, which states that trust congruence enhances social exchange. However, when ethical leadership is weak, this social exchange can result in _unethical pro-supervisor behavior_. Following the four studies, they opened the general discussion section as follows: > “The burgeoning literature on trust congruence indicates that offering employees the trust they desire has beneficial effects when it comes to employee performance. The calibration of supervisors’ trusting behaviors to employees’ actual desire for trust is thus portrayed as an effective leadership style. In this article, we proposed that trust congruence may also have adverse, unintended consequences in the form of unethical pro-supervisor behavior.” ### Why Matching Trust Can Encourage Misconduct What explains this finding? The authors argue that managers create a strong relational bond when they calibrate their trusting behavior to match their employees’ desires. Employees then feel valued, competent, and safe. According to social exchange theory, people then feel obligated to reciprocate. However, without explicit ethical guidance, this reciprocity can manifest as unethical loyalty, such as lying or withholding negative information to protect the supervisor. Moral disengagement theory explains how individuals reframe such acts as justifiable. For example, they might say, “I’m just protecting my boss,” rather than “I’m deceiving the client.” In this way, trust congruence provides the relational energy, while ethical leadership provides the moral direction. ### Implications for Leaders These findings provide valuable guidance for managers and organizations. Trust remains indispensable because it promotes initiative, motivation, and psychological safety. However, this study shows that trust without an ethical foundation can lead to collusion. Therefore, leaders should accompany empowerment with explicit moral framing. When delegating authority, supervisors should clarify the boundaries of responsible autonomy — what can be decided independently and what must be disclosed. Ethical leadership does not mean constant control; instead, it means visibly modeling integrity by communicating moral expectations, rewarding transparency, and addressing wrongdoing consistently. In such environments, employees learn that reciprocating trust means acting responsibly and not concealing mistakes. Organizations should recognize that employees have different preferences regarding the balance between autonomy and structure. Trusting those who want guidance too much can make them feel abandoned, while monitoring those who value independence too closely can make them feel distrusted. Thus, balancing trust requires attentiveness to individual preferences and ongoing ethical dialogue. At a broader level, this research reminds us that _empowerment and ethics are co-dependent_. Trust energizes, while ethics provides orientation. Without ethics, trust can be misguided by loyalty rather than integrity. ### Final Thoughts The promise of empowerment can obscure its risks. This research shows us that good leadership means more than delegating authority; it also means providing ethical direction. Leaders who trust their employees unlock energy and reciprocity. However, without a robust ethical compass, that energy may lead to outcomes that organizations — and society — would rather avoid. Before delegating your next big assignment, ask yourself: Have I provided my team members with the trust and ethical guidance necessary for them to succeed with integrity?