AGI and Ethics
Moral status, fairness, accountability, transparency, and the question of what we owe to advanced AI systems and to each other in their presence.
Executive summary
The ethics of AGI sit at the intersection of long-established moral philosophy and brand-new technological capability. Fairness, accountability, transparency, and the question of moral status of advanced systems are now live concerns for engineers, regulators, and citizens.
Fairness and bias
Models reflect their training data and design choices. Bias mitigation is now a regulatory expectation (NIST AI RMF; EU AI Act high-risk system obligations). Documented harms include disparate medical performance, lending discrimination, and content moderation failures across languages.
Accountability
When an AI system causes harm, who is responsible? Developer, deployer, operator, user? Liability frameworks are being developed across jurisdictions. The EU AI Act creates documented responsibilities along the value chain.
Transparency
What can we know about how a system reached a decision? Mechanistic interpretability research (Anthropic, OpenAI, academic labs) is making genuine progress but a complete account of frontier-model internals remains out of reach.
Moral status of AI systems
Whether advanced AI systems deserve moral consideration is debated. Most researchers (Anthropic, DeepMind, academic ethicists) treat it as an open question worth taking seriously rather than settled either way.
Key takeaways
- 01Ethics applies to development, deployment, and use.
- 02Liability is being allocated by law in real time.
- 03Interpretability is improving but incomplete.
- 04Moral status of AI is debated, not dismissed.
Frequently asked questions
Can AI be ethical?
Systems can be designed to behave in accordance with norms; whether they have ethics in the human sense is a deeper question.
Who is responsible when AI causes harm?
The answer depends on jurisdiction and circumstance. The EU AI Act allocates responsibilities along the value chain.