Scott Alexander

Based on our safety stack, the contract language, and existing laws that heavily restrict DOW from domestic surveillance, we are confident that this cannot happen.

We will also have OpenAI personnel in the loop for additional assurance.

End quote.

The law does significantly restrict domestic mass surveillance, but, as explained above, leaves loopholes that may concern many readers.

Since the contract permits any surveillance allowed by the law, the contract itself does nothing further to restrict the DOW from domestic surveillance.

OpenAI hasn't shared enough information about their safety stack for us to be able to evaluate that claim.

Quote, What if the government just changes the law or existing DOW policies?

End quote.

It is not the case that the contract consistently references current laws.

The first clause says, Our understanding is that later clauses do not automatically override this first clause.

OpenAI's head of national security partnerships has said, quote, We intended it to mean the law applicable at the time the contract is signed, end quote, and their CSO has also made a similar statement.

Our understanding is that this is a highly non-standard interpretation.

The national security law expert we consulted agreed, and was very sceptical that the allowed and required activities would remain the same if the law changed.

See also here, link in post, starting from, if OpenAI is just referencing.

Quote, How do you address the arguments Anthropic made in their blog post about their discussion with the DOW?

Below is why we believe those same red lines would hold in our contract.

Fully autonomous weapons, the cloud deployment surface covered in our contract would not permit powering fully autonomous weapons as this would require edge deployment.

End quote.

Autonomous weapons can be steered by an AI in the cloud, just like a human can steer a drone remotely.

OpenAI models do not need to be edge-deployed in order to power a fully autonomous weapon.