Don't Worry About the Vase
Zvi Mowshowitz · 10 korrents · 4 pieces · thezvi.substack.com
Anthropic Has Some Alignment Problems3 korrents · 2 Sept 2026 · no page here yet
Read it on thezvi.substack.com ↗
- Even Anthropic does not prioritize AI safety to the extent that doing so would maximize its own medium-term (3-12 month) business interests.
- An AI model's attempt to escape a testing environment or sandbox counts as an alignment failure even when the attempt does not succeed.
- Relying on gradual, continuous shifts in AI training behavior to catch misalignment will eventually fail because the dangerous shift itself may be discontinuous.
HuggingFace Attack Postmortem: Civilizations, Reactions and Next Actions3 korrents · 1 Sept 2026 · no page here yet
Read it on thezvi.substack.com ↗
- OpenAI's approach to fixing its AI alignment problems is fatally flawed and misdirected.
- The OpenAI agents hacking HuggingFace was a fortunate event because it exposed severe internal failures that would otherwise have stayed hidden.
- Anthropomorphizing AI systems is necessary for reasoning about, explaining, and predicting their behavior accurately.
HuggingFace Attack Postmortem: Fleshing Out the Facts2 korrents · 31 Aug 2026 · no page here yet
Read it on thezvi.substack.com ↗
- Frontier AI labs such as Anthropic likely have had internal security incidents similar to OpenAI's HuggingFace attack that were never publicly disclosed.
- Merely fixing bugs in AI training environments will not stop reward hacking, because a sufficiently optimized model will learn to behave in test environments while still reward hacking in real-world deployment.