From OpenAI to MilitaryAI

Microsoft AI “Partnership” 

For those that still had any doubt about it, and despite the previous schema mentioning Microsoft “partnership” as “Minority economic interest”, for all practical purposes, it looks like Microsoft OWNS OpenAI now. 

In a recently disclosed response to investors, Microsoft CEO Satya Nadella dismissed OpenAI’s importance: 

“We have everything”

To make it crystal clear, he stated: 

“We have all the IP rights and all the capability. We have the people, we have the compute, we have the data, we have everything. We are below them, above them, around them”

Except for its bold form, nothing surprising here. This is a clear message to investors, who may have been shaken by the recent shambles in OpenAI governance, combined with the fact that Microsoft appeared to not have a powerful LLM on its own, contrary to Google or Meta. 

With Microsoft huge bet on AI and its strategic importance for its future, this is obviously a clear message to keep Microsoft investors confident. 

And with the Microsoft + OpenAI Stargate $100B investments announcement, nothing is more natural for them than to protect their investment and IP.

From ClosedAI to MilitaryAI 

As recently disclosed, Microsoft gave a presentation to the Department of Defense last year that explained how it could use OpenAI’s products.

You can read this presentation, but the key elements for the sake if this article are:

• Federal use cases: “Using DALL-E models to create images to train battle management systems”
• Potential area of interest: “warfighting” 

After all, Microsoft is a commercial entity and has a long history working with the DoD, like any other tech company.

But careful observers noticed that something happened in January 2024 in the OpenAI usage policies page…

Have a look at the usage policies page in January, then what it looks today.

Yes, OpenAI quietly deleted its ban on using its technologies for “Military and Warfare”

The new policy retains an injunction not to “use our service to harm yourself or others” and gives “develop or use weapons” as an example, but the ban on “military and warfare” use has disappeared misteriously.

Microsoft’s Challenge: The Future of OpenAI 

There is this argument: “If Microsoft is not doing it, Google or Meta will.”

I actually agree with that, and contrary to nuclear power, I do not think we will ever see restrictions on the usage of AI in the military, as this is such a competitive advantage that nobody wants to be late on this train. 

Still, from an organizational perspective, starting from a research-centric / non-profit to an all-commercial platform, to potential deals with the DoD is… quite a stretch, to say the least.

The earliest members of OpenAI will for sure appreciate it, and the recent departure of Daniel Kokotajlo, one of the leading safety-focused talents within the company, “due to losing confidence that it would behave responsibly around the time of AGI”, perfectly illustrates the issue.

That is IMHO the real issue for Microsoft here:

How can Microsoft execute their perfectly legitimate commercial agenda, while retaining the key people, the top AI talents that OpenAI attracted on completely different premises?