The "safety" guardrails in Anthropic and OpenAI models are becoming a noticeable problem for security work. And, the reason I'm keeping my Kimi subscription even though it's not a great deal; Kimi subscriptions are quite stingy for the price, but K3 will do vulnerability analysis and make a PoC without requiring you to be on the approved list of Fortune 500 or government entities that have access to Mythos or Daybreak.

I'm not touching Grok. But there are alternatives to Anthropic and OpenAI that don't refuse to do security work.