CA’s Newsom Issues AI Order Covering Watermarks and More

Photo Credit: Steve Johnson

Let the federal v. state artificial intelligence battle continue: Following the publication of the White House’s National Policy Framework for Artificial Intelligence, California Governor Gavin Newsom has signed a multifaceted AI executive order.

Governor Newsom issued that order yesterday, not long after the Framework introduced a variety of “legislative recommendations” at the federal level. Unsurprisingly, given the timing, there’s a political-posturing angle to California’s AI rules.

(The White House recommendations, we reported earlier in March, emphasized an objective of establishing “a federal AI policy framework” to “prevent a fragmented patchwork of state regulations that would hinder our national competitiveness.”)

That’s clear enough from the corresponding press release’s title – “As Trump rolls back protections, Governor Newsom signs first-of-its-kind executive order to strengthen AI protections and responsible use.”

Throw in the long list of rightsholder complaints against leading developers – most recently, BMG levied infringement claims against Anthropic – and it’s safe to say that the AI regulatory landscape contains more than a few moving parts.

Keeping the focus on California’s fresh regulations here, the order instructs the Department of General Services (DGS) and the Department of Technology (CDT) to recommend “new certifications” for “state contracting processes” involving AI developers.

Among other things, said certifications will address safeguards to prevent misuse involving “illegal content” including sexually explicit deepfakes; “harmful bias”; and violations “of civil rights and civil liberties such as free speech, voting, [and] human autonomy.”

Additionally, the Government Operations Agency, “in consultation with DGS and CDT,” has 120 days to propose “any reforms to contractor responsibility provisions” along the same lines as those highlighted above.

More directly related to the federal v. state regulatory showdown: CDT’s information-security officer will “review any new designations of companies as supply chain risks by the federal government.”

And if the official “concludes that the designation is improper,” DGS and CDT “will jointly issue guidance ensuring that departments and agencies can continue to easily procure from that company.”

In closing, the order also calls on the agencies to establish state “employee access to vetted GenAI tools for general use cases with appropriate privacy and cybersecurity safeguards.”

Most interestingly for copyright holders: The executive order’s instructed development of “best practice guidance for departments and agencies to appropriately watermark” all “AI-generated or significantly manipulated images or video.”

Longer term, making AI watermarks the norm across the board would, of course, be a sweeping step – and a step with multiple considerations beneath the surface. How would doing so affect the growing collection of creations incorporating, but not solely resulting from, generative AI? And where will the “significantly manipulated” line be drawn?

Time will tell, but it’ll be worth tracking AI regulatory developments – including in Europe, where several AI legal disputes are unfolding – moving forward.