Anthropic's GLM-5.3 exploit warning sparks accusation of anti-open-weight fear-mongering

ziv_ravid · x · 2026-09-30

Anthropic published an analysis warning that GLM-5.3 can autonomously build working exploits and that its weak safeguards could hand that capability to attackers. Critic zivravid calls this out as hypocrisy: five months ago Anthropic said its own model was too dangerous to release and claimed such capability would inevitably spread — and now that it has, they're allegedly aiming the fear at open-weight competitors. He offers two readings: Anthropic celebrating open models catching up, or nudging governments to intervene, and says he believes the latter.

Related event: Anthropic's GLM-5.3 Cyber Capability Report Sparks Open-Source Safety Debate(22 posts)→

Original post →

More from Models

Models channel →