Regardless of the reason, Google rescinding the promise not to develop AI weapons is concerning

First reported by The Washington Post (via The Verge), it seems Google has made a significant, and concerning, change to its AI principles: removing key language that previously committed them to not developing AI for use in weapons, surveillance that violates international norms, or technology that contravenes human rights. Regardless of the intention of this policy adjustment, this move raises serious questions about Google’s future direction in the rapidly evolving AI space.

The original principles specifically prohibited AI applications in areas that could cause or were likely to cause overall harm. These included the development of AI for:

• Surveillance that violates internationally accepted norms.
• Weapons.
• Technologies whose purpose contravenes widely accepted principles of international law and human rights.

These explicit commitments have been replaced with broader, less specific language focusing on “innovation,” “collaboration,” and “responsible” AI development. While “responsible” development sounds great, the lack of concrete definitions and commitments in the revised principles leaves room for interpretation – and potentially, for applications that were previously not in the conversation.

A blog post by DeepMind CEO Demis Hassabis and Google’s senior executive for technology and society, James Manyika, cites the “complex geopolitical landscape” and “global competition” in AI as driving factors behind these changes. They argue that democracies should lead in AI development, guided by values like freedom and human rights. However, the removal of those former prohibitions raises concerns about how these values will actually be applied.

On top of this policy change, it is notable that Google’s prior involvement in military contracts – such as Project Maven (analyzing drone footage with AI) and Project Nimbus (a cloud contract with the Israeli government) – has already created a bit of internal debate. This policy change suggests a potential further shift towards embracing AI applications with possible military or surveillance implications.

AI clearly has immense potential, but it also carries significant (and seemingly obvious) risks. Regardless of true intent, here, this all definitely feels unsettling to say the least. I don’t want to help create a bunch of unfounded fear with AI, but I don’t think I’m alone in saying this policy change is concerning for sure. How AI is used for weapons, surveillance, and other gray areas for general human rights is a discussion that needs to stay front and center as this all continues to develop at a rapid pace. This particular move by Google (and other AI creators, too) is not a good look, though; that much is certain.