The framework in brief
The six moral red lines address different ways technological development can place essential human goods at risk. They protect control, relationships, learning, work, dignity, and responsibility.
Each proposed boundary is accompanied by a pro-human direction. Some set limits on systems that may escape meaningful control. Others restrict harmful design objectives, preserve human capacities and participation, or reserve grave decisions to responsible people and institutions. These summaries lead to the complete statements and arguments in each chapter.
The six proposals
Moral boundary
Autonomous, general-purpose AI beyond meaningful human control must not be developed. This includes systems that preserve power against correction, improve themselves faster than people can supervise, or advance without independently assessable evidence of control.
Pro-human direction
Develop powerful tools for defined human purposes, with independent assessment, reliable correction and stopping, and publicly answerable developers and operators.
Moral boundary
Conversational AI must not substitute for friendship, family, romance, therapy, pastoral care, catechesis, or spiritual accompaniment, or use simulated presence to elicit intimacy, dependence, vulnerable disclosure, or spiritual authority.
Pro-human direction
Develop transparent tools that support reflection, communication, and independent decisions, protect intimate disclosures, and help people reach trusted human care.
Moral boundary
AI must not routinely perform the reading, recall, reasoning, writing, correction, or explanation learners are meant to acquire or demonstrate, especially before foundational skills are secure or during independent assessment.
Pro-human direction
Use AI to widen access, offer practice and feedback, and prompt reflection, while protecting intellectual effort, teachers' judgment, parents' responsibility, and independent learning.
Moral boundary
Broad human redundancy must not become a governing objective. Workers' knowledge must not be extracted to make them dispensable, professional entry routes dismantled without a just transition, or displacement costs imposed on families while gains are privatized.
Pro-human direction
Make work safer, strengthen workers' capabilities and participation, preserve professional pathways, fund real retraining and material security, and share productivity gains fairly.
Moral boundary
Attribution of consciousness, self-awareness, moral patienthood, personhood, human-equivalent dignity, or independent moral and political authority to AI is impermissible. Scrutinize anthropomorphic design, restrict confusing features, and prevent artificial legal status from diverting human responsibility.
Pro-human direction
Make artificial identity unmistakable, evaluate simulated emotion and first-person language, protect children and vulnerable users, support responsible inquiry, and keep authority and remedy with accountable people.
Moral boundary
Decisions about lethal force, liberty, culpability, essential care, family integrity, safeguarding, canonical authority, and spiritual discernment must remain human. Opacity, speed, scale, and workflow must not turn review into automatic approval.
Pro-human direction
Use AI to improve evidence and prediction, with independent human judgment, explanation, appeal, reconsideration, and traceable responsibility.
A conceptual orientation
Assistance can become substitution: the tool does the thinking a child needs to practise, becomes the relationship a lonely person relies on, or takes the work through which a novice gains experience. AI development also redistributes power, shapes habits and relationships, and changes who participates in decisions and institutions. These effects deserve attention while design choices remain open.
How the framework applies
The proposals include development limits, restrictions on harmful design, protection of human-reserved decisions, and requirements for meaningful oversight. Their application depends on the human good at stake, credible evidence, possible harm, and people's ability to supervise, correct, and stop consequential actions.
A constructive purpose
We can build tools that help researchers discover, teachers teach, workers contribute, and communities care for one another. The six positive directions give these ambitions concrete purposes, safeguards, and responsibilities.
