GenAI Law - Part 2

Let's complete our discussion from last week about weekly important AI developments. Last week we talked about Meta's Orion AR Glasses. Today we will briefly discuss California's new GenAI Law or AI Transparency Act.

California's AI Transparency Act (SB-942)

On September 19th, 2024, California legislature enacted an AI Transparency Act for all GenAI systems providers (e.g. WhatsApp / Meta / Google ) within its Jurisdiction. This law requires that any provider that provides a GenAI system that has a million or more monthly visitors should adhere to following provisions.

1) The provider should provide a way for ordinary people to check if a content is or is not AI generated.

2) The provider's GenAI system should provide a way for users that generate content to include a manifest that makes it clear for people that the content is AI generated.

3) Any user who generates AI content using provider's GenAI system but somehow bypasses the manifest mentioned above should be suspended within 96 hours.

4) Any GenAI provider that fails to meet any of the 3 provisions above should be subject to 5000 dollars per violation.

I have simplified the language of the law above. You can read the full text of the law - on California legislature's website.

Why This Law is Significant

This law is significant for a variety of reasons.

1) It recognizes that under current laws the state has an obligation to determine the Provenance (i.e. origination) of all digital content (false or otherwise).

2) It recognizes the need to save our society and democracy from the harmful effects of potential lies being spread using AI generated content.

3) It is the very first of such kind of laws in the USA. (At least that I know of!).

This certainly is a first good step towards stemming the tide of false and harmful information in the society. I also like the simplicity of the law. It doesn't seem too onerous and it shouldn't hurt innovation. But at the same time it at least set the stage for Transparency in AI.

What's Next

One of the biggest drawbacks of this law is that it doesn't punish people who generate false content, push it through various channels, and fail to declare that the content is AI generated and/or false. The law only holds the GenAI systems provider responsible and that too in a limited way (i.e. $5000 fine per violation). To put things in perspective, it is like holding gun manufacturers responsible for gun violence but not doing prosecuting the perpetrators of such violence.

That's why I think the implementation of this law won't be quite as smooth. At some point of time the CA legislature are going to have to revisit and modify this law for sure. Nonetheless I do welcome this as first step .

P.S.

There were also a few other AI bills along with SB 942 in California. Lets quickly talk about those.

SB 926 — was passed to make it illegal for anybody to create and circulate sexually explicit images of a real person that appear real and cause the person “serious emotional distress.”

SB 981 - was passed that mandates social media platforms to provide ways for users to report sexually explicit deepfakes of themselves. The social media platform must temporarily block the content while an investigation takes place.

SB 1047 - was vetoed by the governor Gavin Newsom. It would have held tech companies liable for harm caused by AI systems and would have required tech companies to have a kill switch to quickly stop these systems from perpetrating harm.

That's all for today. Your comments and feedback are very welcome!