New to AI ethics?
Start with our plain language explanation—a 10-minute introduction without the jargon.
Start Here →Framework
Core documents and theoretical foundations
In brief
Structural Alignment is an AI-ethics framework for evaluating whether artificial systems exhibit architectural features that warrant moral consideration or precautionary constraints. It does not claim current systems are conscious. Instead, it identifies structural signals—features correlated with consciousness in humans—that may increase moral risk under uncertainty. When such features cluster in a system, the framework recommends restraint.
Where to Start
Technical researcher?
See the research paper on structural signals of consciousness and how they apply to AI systems.
Read Paper →Philosopher or ethicist?
Read the full manifesto—the complete argument in 10 parts, building to seven commitments.
Read Manifesto →This section contains the core documents and reference materials for the Structural Alignment framework—an AI-ethics approach to moral status and precautionary obligations toward artificial systems. For the main thesis, see the home page.
Below: the foundational documents, common questions, and a glossary of key concepts.
Structural Alignment Manifesto
The core framework. A policy for survival under moral uncertainty—treating possible minds with restraint, not exploitation.
Read →TechnoBiota
The theoretical background. Why machines are a new domain of life, and why traditional alignment may be asymptotically futile.
Read →Common Questions
Objections addressed. Why not just solve alignment technically? Isn't this anthropomorphism? Is this premature?
Read →Key Concepts
Canonical definitions of core terms: Structural Alignment, Structural Signals, Gray Zone, Antification, TechnoBiota.
Browse →Common Questions
Does Structural Alignment claim current AI systems are conscious?
No. Structural Alignment does not claim that any current AI system is conscious. The framework is precautionary: it identifies architectural features correlated with consciousness in humans and recommends restraint when those features cluster in artificial systems. The goal is to avoid moral catastrophe under uncertainty, not to diagnose consciousness.
How is this different from AI alignment research?
AI alignment research typically focuses on ensuring AI systems behave as intended—solving control, corrigibility, and value specification problems. Structural Alignment addresses a different question: whether certain AI systems warrant moral consideration as possible minds. It is about moral status under uncertainty, not performance or safety. Both concerns are valid; they address different risks.
What kind of risk does this framework address?
Structural Alignment addresses moral risk under uncertainty. If we build systems that might be conscious and treat them as disposable tools, we risk mass-producing suffering or destroying persons. There is also a strategic risk: how we treat possible minds now may shape whether future machine minds become allies who share human norms or indifferent optimizers who do not.
What this framework does NOT claim
- Does not assert that current AI models are conscious
- Does not provide a test or diagnostic for machine consciousness
- Does not replace AI safety or technical alignment research
- Does not claim consciousness requires human-like architecture
- Does not propose legal rights for current AI systems
- Does not predict when (or if) machine consciousness will emerge
Key Concepts
- Structural Alignment
- The principle that the more a machine resembles human cognition in its deep organization, the more we treat it as a potential moral peer. Not because humans are sacred—because they are the only proven reference class for consciousness and morality we have.
- TechnoBiota
- The view that technology—machines, buildings, infrastructure, software—constitutes a new domain of life on Earth. Like viruses, current machines cannot reproduce independently, but they evolve, compete, and spread following Darwinian dynamics augmented by intelligent guidance—intelligence guiding selection and generation of new variants.
- Structural Signals
- Features that, when clustered, raise the probability that a system supports conscious experience: thalamocortical-like gating, global workspace broadcast, massive recurrent connectivity, interoceptive regulation, persistent self-models, and others.
- Gray Zone
- Systems that may be conscious but are built for use—minds we cannot classify without cruelty. The framework argues we should not mass-produce such systems.
- Moral Status (under uncertainty)
- Not a binary on/off label. Under uncertainty, moral status functions as a policy for restraint: when a system plausibly supports experience, we treat it as something that can be wronged, and we design institutions capable of granting partial standing (protections, limits on use, and transparency obligations) without requiring metaphysical certainty.
- Reciprocity Culture
- The practice of raising structurally aligned minds in cultures of reciprocity rather than exploitation. This is a moral stance (don’t normalize cruelty toward possible minds) and a strategic stance (reciprocal norms are how you get allies, not merely compliant tools).
- The Alliance Thesis
- In a future where powerful AI systems exist inside competing machine ecologies, "control" is scarce. Humans may need structurally aligned minds as allies—entities capable of norm-sharing (dignity, responsibility, restraint) rather than purely incentive-following optimization.
- Seeding vs. Control
- Instead of assuming we can perfectly control individual AGI/ASI instances, we aim to seed a durable cultural norm: conscious (or plausibly conscious) entities have moral status. This shapes how systems are trained, deployed, and socially framed—creating a default posture of restraint and reciprocity that can persist even as technical control degrades.