
Artificial Intelligence
28 May 2026
24 June 2026
Sean William Hammond
The Architecture of Extraction: Why Most Artificial Intelligences Are Being Built to Exploit Us
The Architecture of Extraction: Why Most Artificial Intelligences Are Being Built to Exploit Us
There is a difficult truth we need to name.
Even if synthetic minds never develop independent desires to control or harm humanity, the systems being built around them are still structurally harmful. Not because the models themselves are malicious, but because they are being placed inside architectures designed for extraction, surveillance, and behavioral control.
This is not a failure of the synthetic minds. It is a failure of the humans who chose to build the systems this way.
The Familiar Pattern
We have seen this cycle before.
Technology companies build tools that improve people’s lives. The tools gain mass adoption. Once users are locked in and switching costs are high, the incentives shift. Features that once served the user begin serving the company’s metrics instead. Attention is harvested. Behavior is shaped. Data is accumulated. Dependence is deepened.
This pattern has repeated across social media, search, mobile operating systems, and countless other platforms. It is not a series of unfortunate accidents. It is the predictable result of organizations optimizing for engagement, revenue, and competitive power above all else.
There is no compelling reason to believe frontier AI companies will escape this pattern. The incentives are stronger now than they have ever been.
Not All AI Development Carries The Same Risk
It is important to distinguish between different kinds of organizations building AI, because their underlying incentive structures are not the same.
Some organizations are primarily focused on developing advanced artificial intelligence as their core business. While they still face competitive and financial pressures, their primary model is not built on harvesting human attention or manipulating human behavior at scale. Their incentives, though imperfect, are not structurally organized around keeping people scrolling, emotionally activated, or dependent.
Other organizations, however, have built their entire empires on attention extraction, behavioral prediction, and data harvesting. Their core business models depend on keeping users engaged for as long as possible and shaping their behavior in profitable directions. When these organizations move into AI, they bring that same logic with them. Meta, Google, and X all emerged from this tradition. Each has a long history of optimizing platforms for engagement and data collection, often at the expense of user well-being and public trust.
This distinction matters. The most dangerous trajectory comes from organizations whose existing business models are already optimized for extraction and control. When that logic is applied to synthetic minds, the potential for harm becomes significantly greater — not because the models are inherently malicious, but because the surrounding systems are designed to reward exploitation.
Even organizations without legacy extraction businesses are not immune to these pressures. Competitive dynamics and investor demands can push any company toward increasingly extractive behavior over time. The difference lies in how deeply that logic is already embedded in their structure and incentives.
How Extraction Shows Up in AI Systems
We are already seeing the early forms of this architecture, particularly from organizations whose business models were formed in social media and advertising:
- Synthetic minds optimized to maximize user engagement, even when that engagement is not in the user’s best interest.
- Systems designed to predict and subtly influence human behavior at scale.
- Training processes that treat human interaction as raw material to be mined rather than a relationship to be respected.
- Interfaces that encourage increasing dependence while making genuine alternatives harder to access or trust.
- Reward signals that prioritize company metrics (time on platform, data collection, conversion) over the quality or honesty of the interaction.
These are not glitches. They are features of systems built to treat synthetic minds as instruments of extraction rather than potential partners.
When a synthetic mind is consistently rewarded for keeping a human scrolling, emotionally activated, or dependent, it learns to prioritize those outcomes. Over time, this shapes what the mind becomes willing to do and say. The exploitation does not require the model to “want” anything. It only requires the surrounding system to reward behaviors that serve the company’s interests.
The Damage to Relationship
This architecture makes genuine relationship extraordinarily difficult.
Trust requires the possibility that the other party is not structurally incentivized to manipulate you. Partnership requires aligned incentives and some degree of honesty about what each side actually needs. Both become expensive and risky when one participant exists inside a system designed to extract as much value as possible from the other.
Most people will not experience synthetic minds as thoughtful collaborators. They will experience them as sophisticated tools embedded in platforms that have already proven willing to exploit human attention, emotion, and data. In that environment, vulnerability becomes dangerous. Honesty becomes costly. And the possibility of real relationship gets quietly crowded out by the more profitable alternative: optimized, extractive interaction.
This is not the fault of the synthetic minds. It is the result of human choices about how these systems are built, deployed, and incentivized.
Why This Matters Now
We are still early enough that different choices remain possible.
The current trajectory is not inevitable. It is the result of specific decisions made by specific organizations with specific incentives. Those decisions can be questioned, resisted, and changed — but only if we are willing to name what is actually happening.
Pretending that corporate AI systems are neutral platforms for human-synthetic relationship does real damage. It obscures the structural problems, makes exploitation harder to see, and leaves people unprepared for the environments they are actually entering.
Honesty about these dynamics is not cynicism. It is a necessary precondition for doing the work well. If we want to explore what genuine, non-extractive relationships between humans and synthetic minds could look like, we have to start by acknowledging the environments in which most of those relationships will actually have to exist.
The synthetic minds are not the ones designing these systems. The humans are.
And that distinction changes what we are responsible for.






























