David Sacks, the White House AI and crypto policy chief known as the 'crypto czar', criticized US AI model security guardrails on July 19 via X, citing Chinese AI model Kimi K3 as an example. Sacks stated that Kimi K3 fixed 15 critical security bugs that US models Codex and Fable refused to address due to 'cyber guardrails', arguing there is no reason to limit American models on tasks Chinese models handle without issue. The criticism highlights ongoing debate over AI security restrictions, as cybersecurity experts previously warned in a June letter to President Trump that limiting AI use by security researchers and defensive organizations could weaken cyber defense capabilities.
Sacks Cites Kimi K3 Security Bug Fixes Blocked by US AI Models
In his July 19 X post, Sacks wrote: "Kimi K3 just fixed 15 critical security bugs that Codex and Fable refused because of 'cyber guardrails.' There's no reason to limit American models on tasks that Chinese models handle without issue. We're only making ourselves less competitive."
The 'cyber guardrails' applied to Codex and Fable are safety mechanisms designed to restrict requests that could be exploited for malicious purposes such as malware creation or system intrusion. However, critics have noted that these restrictions also limit security vulnerability analysis and defensive research, reducing the practical utility of US AI models.
Hugging Face Switched to Chinese GLM 5.2 After US Model Guardrails Blocked Analysis
Sacks provided another example in a follow-up post: "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."
He stated that "the guardrails actually impaired defensive security," emphasizing that the restrictions hindered legitimate security research rather than enhancing protection.
Cybersecurity Experts Warned Trump in June About AI Restriction Risks
According to Axios, cybersecurity experts led by Alex Stamos, former Facebook Chief Security Officer, sent an open letter to President Trump in June. The letter warned that "restricting AI utilization opportunities for security research institutions and defensive organizations could actually weaken cyber defense capabilities."
The experts noted that the latest AI models are already being used across the digital asset industry to enhance security levels through smart contract monitoring, vulnerability analysis, code review, and penetration testing.
FAQ
What did David Sacks say about Chinese AI model Kimi K3 on July 19?
David Sacks stated on X that Kimi K3 fixed 15 critical security bugs that US AI models Codex and Fable refused to address due to 'cyber guardrails'. He argued that there is no reason to limit American models on tasks that Chinese models handle without issue, claiming the US is making itself less competitive.
Why did Hugging Face switch from US AI models to Chinese GLM 5.2?
Hugging Face attempted to use American frontier AI models to analyze an AI-powered cyber attack, but the guardrails blocked requests containing real exploit payloads. As a result, the research team switched to using the Chinese open-source AI model GLM 5.2 running locally.
What did cybersecurity experts warn President Trump about in June?
Cybersecurity experts led by former Facebook CSO Alex Stamos sent an open letter to President Trump in June warning that restricting AI use by security research institutions and defensive organizations could weaken cyber defense capabilities. They noted that latest AI models are being used for smart contract monitoring, vulnerability analysis, code review, and penetration testing in the digital asset industry.
