'Cyber Guardrails' Fight Goes Mainstream: Kimi K3 Fixed 15 Security Bugs Codex and Fable Refused, and David Sacks Amplified It
A developer reported that both OpenAI's Codex and Anthropic's Fable refused, citing cyber guardrails, to fix a report full of security issues in their own codebase, while Moonshot's Kimi K3 fixed all of them; White House AI advisor David Sacks reposted it with a competitiveness frame — 'There's no reason to limit American models on tasks that Chinese models handle without issue.' Hugging Face CEO Clément Delangue replied that his team hit the same wall the same week: 'Very scary to be guardrailed as a defender when you know attackers are likely bypassing.' The session has not been independently reproduced and severity labels are unvalidated, so treat this as evidence that hosted guardrails can block authorized fix generation — not as a CVE drop.
↳ Follow the thread