Export Controls on AI Models Like Fable 5 Undermine US Cyber Defense, Expert Argues

Summary

The article criticizes export control directives targeting AI models like Fable 5 and Mythos, arguing they harm US cyber defense capabilities while failing to impede attackers. The author claims to be one of the few outside experts who has actually read the underlying research paper, and explains that the study used open-source code with known vulnerabilities plus deliberately planted flaws to test AI code review capabilities. When Fable 5 refused to participate, it highlighted the unintended consequence of export controls weakening defensive cybersecurity tools without affecting malicious actors who already have access to the technology.

Source

bskyExport Controls on AI Models Like Fable 5 Undermine US Cyber Defense, Expert Argueslutasecurity.com