This is educational material and does not constitute legal advice nor is any attorney/client relationship created with this article, hence you should contact and engage an attorney if you have any legal questions. No warranties, express or implied, are made with respect to its accuracy. Information contained herein, or information relied upon, is subject to change without notice.

Everything you need to know about California’s new AI law

• On September 29, 2025, California enacted SB 53 (Transparency in Frontier Artificial Intelligence Act) — the first U.S. law to mandate safety and transparency rules for frontier AI. 

• It applies only to “large frontier developers” that exceed 10^26 FLOPS compute use + $500M annual revenue — e.g. OpenAI, Google, Meta. 

• Regulated firms must publish a frontier AI framework and transparency reports. They must also confidentially report “critical safety incidents” to California’s Office of Emergency Services. 

• Whistleblower protections cover employees, contractors, vendors, and board members who disclose catastrophic risk or legal noncompliance. 

• Beginning in 2030, those firms must undergo third-party audits to confirm compliance with published frameworks. 

• The California AG enforces SB 53, with penalties up to $1M per violation. The law emphasizes enforcing transparency requirements rather than penalizing harms directly.

Sora 2 is here, and I got in; invite 5 people to subscribe to Future Perfect to get your access code!

🎉 Sora 2 is out, and I'm in!

Want an access code? Grab the app, follow me (username = autonomoushg), make sure you’re subscribed, and refer 5 new subscribers to me!

⚠️ First things first: login is via your ChatGPT account which of course makes sense; what doesn't make sense is that, according to Paul Yacoubian, if you try to delete your Sora account, apparently it wipes your entire ChatGPT account too (presumably because they're linked). Not sure if this is a bug or a feature.

So there's that.

Also, I took a look at OpenAI's latest Terms of Service and Privacy Policy so you don't have to. Surprisingly, nothing's been updated since June 2025 (Privacy Policy) and December 2024 (Terms of Use); I would have expected a major update following the Sora 2 release.

So this is just a refresher covering the good, the bad, the "huh?!" and action items for you to take:

✅ Good
• You own your inputs and outputs (but remember you can't copyright AI-generated output)
• API/Business/Enterprise data isn’t used to train models by default but…

🚨 Bad
• …Consumer ChatGPT chats train models by default unless you opt out
• Broad device and location collection
• Mandatory arbitration + class-action waiver, 30-day opt-out
• Business/Enterprise admins can access employee content if employee account is via business email domain
• Court order still standing for OpenAI to preserve consumer ChatGPT/API content, even deleted chats, until lifted (3rd party services like GC AI which have negotiated ZDR policies seem to be exempt)

🤔 Huh?
• OpenAI says that copyright holders will need to "opt out" if they don't want their work to appear in Sora-generated videos.
• For all the arguments I've made about how AI-generated output doesn't (in general) infringe on copyright this is NOT how copyright works.

💡Action Items
 • Personal accounts: Settings → Data Controls → turn off “Improve the model for everyone” and “Include audio recordings” and "Include video recordings"
• Business: Use API/Enterprise. If you need logs minimized, request and negotiate ZDR for eligible endpoints
• Oh and, yeah, probably opt out if you don't want Sora copying your work.

Looking for past newsletters? You can find them all here.