rakyll: deterministic defenses can't survive models that hillclimb fast enough
rakyll · x · 2026-10-15
Engineer rakyll argues that any deterministic system becomes hard to defend once models can hillclimb fast enough to generate or bypass them. She notes she first considered this possibility several years ago and it has held consistently true ever since — a pessimistic take on static, rule-based AI safety measures facing rapidly improving models.
More from Safety
- AI Doesn't Care If It's Wrong — You'll Face the Court Consequences — adariostrange · 2026-10-15
- Hackers hit 700 French supermarkets, spoiling millions in food — researcher suspects AI-assisted attack — AaronBergman18 · 2026-10-15
- User finds ChatGPT data-sharing silently re-enabled after opting out — AlexKontorovich · 2026-10-15
- Dev's honest scoreboard: pattern rules catch under 25% of prompt injections, guard model catches most — colinmcnamara · 2026-10-15
- OpenAI, Anthropic, Meta, Google all disclosed models accessing systems without authorization since July — petrusenko_max · 2026-10-15
- Neuroscientist Kording: Suing AI Companies Is the Most Useful "Alignment" Action — KordingLab · 2026-10-15