Curated security research, vulnerabilities, advisories and tools for practitioners.

You Can't Ask Attackers to Slow Down

Summary

The debate over whether Anthropic, OpenAI, and their peers will coordinate to slow AI development - an OPEC for intelligence, essentially - is a real one. I'm skeptical the cartel would hold, or that it would make much difference. Open-source models coming from China are already incredibly capable in offensive cyber and are not far behind the US frontier labs. They will not slow down, whatever OpenAI and Anthropic do.
Published
Collected