Everyone Missed The Important Clause

Over the weekend, I joined a thought-provoking discussion sparked by a post from former Italian parliament member and computer engineer Davide Zanichelli regarding the growing debate around OpenAI, Anthropic, and US Department of War cooperation.

Much of the public reaction framed the issue as a simple ethics divide. On the surface, one company said yes, and another said no.

Reading the actual agreement matters.

OpenAI’s published framework anchors usage limits to what is considered lawful. At a glance, this sounds reassuring, but it introduces a more consequential question: if capability boundaries are tied to law rather than technical constraint, those boundaries can evolve whenever legal or policy definitions change. This shifts the discussion away from corporate ethics and toward governance custody.

Both Sam Altman and Dario Amodei have signaled willingness to work with governments under defined conditions. The real distinction may not be moral posture, but where control ultimately resides once systems are deployed. If “lawful use” becomes the governing standard, several uncomfortable questions follow:

Who ultimately defines lawful use as AI capability accelerates?

How can citizens verify that humans remain meaningfully in the loop when AI systems influence surveillance, targeting, or operational decisions?

And what governance structure, and qualifications, should be required for anyone authorized to approve AI-enabled actions affecting human lives or privacy?

These are not company questions. They are institutional ones. We may be entering a phase where AI safety depends less on model alignment and more on who holds authority to interpret, authorize, and redefine acceptable use over time.

Original human-authored work with limited AI-assisted drafting and illustration.

Consulting: Need independent analysis or security support? See AI & Cybersecurity Consulting.