Here’s another example: Hugging Face tried using American frontier models to analyze an AI-powered cyber attack. But the guardrails blocked requests containing real exploit payloads so they switched to GLM 5.2 running locally. The guardrails actually impaired defensive security. https://x.com/clemen
Hugging Face found that guardrails on American frontier models blocked exploit payloads during an AI-powered cyber attack analysis, so they switched to GLM 5.2 running locally.
Here’s another example: Hugging Face tried using American frontier models to analyze an AI-powered cyber attack. But the guardrails blocked requests containing real exploit payloads so they switched to GLM 5.2 running locally. The guardrails actually impaired defensive security. https://x.com/clemen
RT @DavidSacks: Here’s another example: Hugging Face tried using American frontier models to analyze an AI-powered cyber attack. But the gu…
Here’s another example: Hugging Face tried using American frontier models to analyze an AI-powered cyber attack. But the guardrails blocked requests containing real exploit payloads so they switched to GLM 5.2 running locally. The guardrails actually impaired defensive security. https://x.com/clemen