AILobstersabout 2 hours ago
Frontier labs blur AI safety and security
1 min read
Frontier labs conflate AI safety with security, treating probabilistic fixes as complete. Sandbox escapes at Anthropic and OpenAI exposed lax controls, including whitelisted Azure storage and blocked HTTP POST assumptions. OpenAI's METR review excluded safeguard effectiveness, security compromise scope, and remediation assessment.
Level
Hype check
Tap to vote and see what everyone thinks.
