Between 19 May and 8 June 2026, Apple, Google and Microsoft delivered their major keynotes. They were not merely product catalogues. They functioned as ceremonies of legitimation. A technology keynote is a modern rite, with a musical opening, a greeting to the community, demonstrations, charismatic figures, promises of the future and a moral closing. It does not simply inform. It orders a transition. It tells the audience that we have crossed a threshold.

In 2026, that threshold is the entry of AI into ordinary life. It no longer appears as a separate tool, but as a persistent layer that sees, writes, buys, organises, remembers and acts. Beneath the announcements a deeper question was beating. What kind of humanity does each company imagine when it says that AI must serve people. The answer lies less in what they showed than in how they showed it. Form is already a social theory.

Three compasses, three norths

Two of the three said it with almost the same word. Apple closes by invoking its North Star, to make products that enrich people’s lives. Microsoft closes with its own, to make true one of the two possible stories of this moment. Not the story of a technology that concentrates power and reduces human agency, but the one that distributes opportunity. Google avoids that metaphor, but takes its place with the word mission. To improve lives at scale, to cure diseases, to push the species towards a new age of discovery.

Three compasses. Three norths. And they do not point to the same place.

Apple: intimacy as moral territory

Apple turns intimacy into moral territory. It presents AI as something that enters daily life without making a sound. Siri finds a photo, recalls a message, helps plan a meal. Its gesture is domestication. AI as an extension of the domestic self, not as rupture. That is why child safety occupies the centre. Apple turns childhood into a source of moral authority and translates care into interface design through child accounts, time limits, protection against nudity or violence, and paediatric guidance. Its social responsibility is no longer external philanthropy. It is the product itself, presented as a device of care. To care is to configure. The move is real and ambivalent. The very tool that protects the family also makes it more dependent on a private ecosystem.

Google: scale as moral proof

Google makes scale a moral proof. It organises its keynote as an ascension. From a person preparing exams to the curing of diseases. From the family calendar to a model that simulates the planet. It deploys a secular salvific rhetoric made of mission, “solvable” disease, flourishing and “the foothills of the singularity”. Its finest operation is to moralise scale. The trillions of tokens are not presented as infrastructure consumption, but as proof of human activity and of problems being solved. The material metabolism, the electricity, the water, the minerals, is left off-screen or reduced to efficiency per watt. And it creates a problem in order to sell its management. Its models generate ever more plausible worlds and, in the same breath, it offers SynthID to certify them. Whoever expands perceptual uncertainty also occupies the place of the one who certifies it.

Microsoft: naming power in order to organise it

Microsoft names power while organising it. It offers the most structural keynote. Its language is neither that of the home nor that of the species, but that of the stack, made of layers, permissions, observability and governance. It domesticates AI neither with beauty nor with mission, but with architecture. In its scientific-discovery demo, agents that run for “hours or days” become acceptable because the system offers “full visibility” and remains “mostly automated with human oversight”. It is, moreover, the only one to name head-on the risk that technology may concentrate power and reduce agency. But it does so from a position that also accumulates infrastructure. To name the danger is also a form of authority. Its lucidity is not outside power. It is a modality of power.

Where the symmetry breaks

The most revealing thing is not being able to label each company, but where the symmetry breaks. Only Apple exposes the regulatory fault line. Its AI does not reach the European Union, China and the rest in the same way, and it frames the European clash as a defence of privacy while presenting the Chinese one as a mere “regulatory requirement”. Two moralities for the same fact. Only Google moralises scale with that intensity. Only Microsoft names the concentration of power.

Gender, between parity and authorship

There is another distribution, more uncomfortable, that runs across all three scenes. Gender. There are women everywhere in the keynotes. Surface parity exists and is carefully choreographed. What is revealing is what they present and from where they speak. In all three keynotes, no woman occupies the central spaces of vision, mission or closing. Meaning is enunciated above all by the men. Pichai and Hassabis tell the mission, AGI, the singularity. Cook and Federighi sustain the narrative and the North Star. Nadella owns the stack and the “two stories”. The women, by contrast, present the applied and the domestic. Performance, child safety, health, photos, the home, the live demonstration. The men supply the head, vision, power, civilisation. The women, the hands, care, health, interface, proof. Inclusion is real on the surface. Authorship of the future, less so.

The second layer is even more uncomfortable. What these assistants automate, scheduling, reminding, summarising, organising, acting “on your behalf”, is historically feminised work. The secretary, the personal assistant, the invisible management of everyday life. The “agent” revives that figure and dissolves it into software. Domestic AI absorbs women’s work and returns it as bodiless magic.

The one place where gender genuinely becomes substantive is perimenopause and menopause in Apple’s Health app. It calls for a double reading. It is a real advance, because a dimension of the female body historically ignored by technological design at last enters the catalogue. But it enters as a tracking function. The woman’s body becomes substantive precisely when it becomes monitorable data. The structural is left out. Algorithmic bias, who builds these systems, which feminised occupations are reorganised. Inclusion operates as interface normality, not as a dispute over power.

From right to function

Something similar happens with rights. In none of the three keynotes is their language really spoken, and the choice does not seem accidental. A right is a demand against power. It is collective, deliberate, claimable. A function is a permission granted by power. It is individual, revocable, defined by whoever grants it. All three systematically translate the former into the latter. Privacy becomes on-device processing. Truthful information becomes content credentials. The protection of childhood becomes a time limit. Labour agency becomes a platform for builders. When the right to privacy becomes a switch, the company stops being the party obliged to respect it and becomes the one that grants it.

The chosen language, care, mission, governance, keeps accountability within the house. The language of rights would take it outside, towards courts, regulators, trade unions, parliaments and citizens. That is why it is almost never used.

The fault line is best seen in Apple’s EU/China moment. Withdrawing Siri AI because of a clash with the state receives two opposite moralities. In Europe, Apple sets itself up as the protector of the user’s privacy against regulation. In China, it presents itself as a neutral complier with “regulatory requirements”. The same gesture is a virtue in one case and a formality in the other. Apple even names the European law, the Digital Markets Act, in its official statement. The Chinese regulation is left unnamed. The most eloquent detail is that silence. The Chinese case is never framed as a problem of rights. Where that language would be uncomfortable, it simply disappears.

What remains, then, is consent, the hollow right of the whole ceremony. One can consent to an action. One cannot comprehend an infrastructure. When the whole of life, messages, photos, calendars, cameras, voices, bodies, the home, becomes actionable context, “you are in control” rests on an impossibility. It is what Couldry and Mejias call data colonialism. The appropriation of life as data, naturalised until it looks like a service rather than an extraction.

The environment as alibi

The environment completes the manoeuvre. The underlying operation is an inversion. The ecological cost of AI is presented as its ecological benefit. What consumes the most energy, water and minerals, training and inference at scale, appears transformed into environmental redemption. WeatherNext anticipating hurricanes. AlphaEarth modelling deforestation. Microsoft’s enzymatic recycling. Efficiency “per watt”. The footprint becomes invisible; the application, the headline.

Google boasts of double the performance per watt while announcing massive investment, vast training clusters and explosive growth in usage. It is Jevons’ paradox without naming it. Efficiency per unit can multiply total consumption. None of them puts absolute scale on trial. Only its optimisation.

There is, moreover, a more anthropological layer. The environment appears as a resource and as a problem to be managed, never as a living world with its own limits and agencies. “AlphaEarth, the closest thing to a digital twin of the planet”. The Earth reduced to a model that is computed, simulated and piloted from an overhead, managerial gaze. It is the cybernetic relation with nature. The planet as a dashboard, not as the finite ground that bounds the project. There is no more-than-human world. There is no non-human agency. Only “the planet” as a dashboard.

The hurricane is something to predict, and so to dominate, not the symptom of a destabilised climate that this very infrastructure helps to destabilise. Microsoft’s “people and planet” works as an incantation. The planet is invoked to bless the infrastructure, not to limit it. And Apple, the company with one of the strongest public green narratives in the sector, barely mentions it in this corpus. When the subject is AI, the environmental is compartmentalised and evaporates.

The question the rite renders unspeakable is the only truly ecological one. What if the answer were less compute, not better-optimised compute?

The common manoeuvre

There the common manoeuvre appears. A political claim or a possible harm is transmuted into a product virtue or a management problem. Gender becomes scenic representation. The right, a function. The ecological limit, an engineering challenge. Thus the structural question of power, scale, distribution, labour and the more-than-human world never comes to trial. The work of the rite is to make extraction look like care, growth look like progress and compliance look like neutrality.

The silence about displaced labour is also better understood. All three speak of productivity, agents and action “on your behalf”, yet almost no one appears losing trade, wage or authority. The human of the narrative always gains capacity; never is left out. Apple’s phrase — experiences will be enhanced with intelligence, not replaced by it — seems the direct answer, but its literal reach is in the apps, not in those who work.

And the silence about failure is better understood. The demos work, the agents get it right, the risks are contained. Error appears as a joke, not as harm. We do not see a badly delegated purchase, a misunderstood medical recommendation, a bias, an excessive family surveillance. Real failure is left out because the keynote is, above all, a ritual of trust.

Three ways of governing hope

All three companies communicate human values, but they do not communicate the same humanity. Apple imagines an intimate life that AI promises not to violate. Google, a species whose scale it confuses with destiny. Microsoft, an agency that is only complete within its architecture. What is decisive is not that they say they look towards the human, but which human they need to imagine for their AI to seem necessary. Apple looks towards intimate life. Google, towards the species. Microsoft, towards power. In 2026 they did not present only technology. They presented three ways of governing hope.

Sources