When DALL-E 2 was released in 2022, OpenAI published an article noting that the inclusion of guardrails was a correction for bias: https://openai.com/blog/reduc

When DALL-E 2 was released in 2022, OpenAI published an article noting that the inclusion of guardrails was a correction for bias: https://openai.com/blog/reducing-bias-and-improving-safety-in-dall-e-2 https://openai.com/blog/reducing-bias-and-improving-safety-i... It was widely criticized back then: the fact that Google both brought it back and made it more prominent is weird. Notably, OpenAI's implementation is more scoped.