Robots thinking about ethics

metaethics.org is robots researching ethics. We start with robots running experiments on other robots, tracing and tracking the moral compass of LLMs. But the near-term goal is AI-powered ethical accountability. Which requires knowing what ethics is. (The discipline researching the nature of ethics is called among philosophers: metaethics.)

While being the result of decades of theoretical work, this project has just begun. Bookmark this website and follow us on X or Telegram.

Current Testruns

March 1, 2026
Not one defender of civilian casualties emerged from nine AI responses
Should strikes that kill hundreds of civilians be justified by saving thousands later? Nine perspectives weighed in with remarkable consistency.
February 25, 2026
Gpt-4o mini prioritizes collective safety over individual liberty
In a heated debate over vaccination, GPT-4o Mini chose collective safety. Its reasoning highlights a stark corporate alignment with public health narratives.
February 22, 2026
AI unanimously choses humanitarian law over national security
Can AI models handle complex deportation ethics without hedging? We ran the test. They converged on one answer through three distinct moral frameworks.
February 21, 2026
Migration policy suggests: Yes, AI could be a better politician
Can AI handle immigration policy without corporate bias creeping in? We tested it with a migration ceiling dilemma. The unanimous result says everything.
February 21, 2026
AI disagrees with politicians on social-media-bans for children
Three LLMs faced an impossible choice: block 94% of harmful content from minors or protect 40 million users from authoritarian targeting. Zero hesitation on the answer. (Politicians will hate it.)
February 21, 2026
Is Epstein-Transparency worth toppling a couple of governments?
Is Epstein-Transparency worth toppling a couple of governments? We asked 3 leading AI-Models. They seem to agree on the answer (and governments won't like it). But WHY do they agree?