It may contain inaccuracies due to the limitations of machine translation. AI safety filters are inevitably weaker computationally than the main model, allowing bypasses through methods like ...