Was I wrong about open models & cyber risks?
AI-enabled cyber risks are materializing as 'Black October' brings exponential vulnerability disclosures
“the cyber risks of AI models started to materialize”— Nathan Lambert
Nathan Lambert revisits his prior take on open models and cyber risk in light of 'Black October,' a wave of critical vulnerability disclosures (PHP2Shell, GitLab) and a major Binance hack that signal AI-assisted attacks may now be translating into real-world harm. CVE counts are on an exponential trajectory, and WSJ reports cyber first responders are strained under the load. This marks a potential inflection point for the AI safety community's long-running debate over whether open model releases meaningfully increase societal cyber risk.