From ZDNet: 2024-04-12 12:28:00

Documents revealed Microsoft Azure pitched its version of OpenAI’s DALL-E image generator to the US military in October 2023. The presentation suggested using DALL-E to train battlefield tools via simulation. This contradicted OpenAI’s historical stance against military use but has since been updated to allow certain defensive applications.

Microsoft presented DALL-E under the Azure OpenAI umbrella, combining cloud computing with the generative AI power of OpenAI. The DoD could use AOAI for various machine learning tasks, from content analysis to creating images to train battle management systems. OpenAI removed the ban on military use from its policies in January, now only prohibiting harm to individuals or property.

The change in policy aims to align OpenAI’s mission with certain military use cases, such as cybersecurity and defensive measures. However, applications like weapons development, harm to others, and property destruction are still prohibited. Microsoft clarified that the examples in the pitch were potential use cases and not implemented. OpenAI denied involvement in the presentation and emphasized their policies against harmful applications of their tools.

Maintaining policies across derivative versions of base technology can be challenging. Microsoft’s established relationship with the US military through previous contracts makes AOAI a suitable option due to Azure’s enhanced security infrastructure. It remains to be seen how OpenAI will navigate the use of its tools in military applications amidst continued collaboration with Microsoft and the DoD.



Read more at ZDNet: Despite DALL-E military pitch, OpenAI maintains its tools won’t be used to develop weapons