Largely because they feared that large language models would be useful for large scale disinformation, election interference, impersonation, automated phishing,

Largely because they feared that large language models would be useful for large scale disinformation, election interference, impersonation, automated phishing, etc. https://openai.com/index/better-language-models/ https://openai.com/index/better-language-models/ See "Policy Implications" and down.