Elon Musk fights California over First Amendment

10 snips

Mar 9, 2026

A legal showdown over California transparency laws and whether forcing dataset disclosures violates constitutional protections. The rise of lawsuits from creators over scraped copyrighted training data and how companies respond. Technical risks like model collapse from training on AI-generated data and the limits of audits and probabilistic testing.

Ask episode

AI Snips

Chapters

Transcript

Episode notes

INSIGHT

Synthetic Data Inbreeding Causes Model Collapse

AI models degrade when trained on generations of AI output, a problem dubbed synthetic data inbreeding that causes bizarre hallucinations like “jackrabbit” tangents.
Hosts describe third- or fourth-generation retraining producing nonsense about jackrabbit breeding instead of historical architecture, showing loss of human anchor.

INSIGHT

Transparency Laws Demand Granular Dataset Disclosures

California-style transparency laws force generative AI developers to post high-level summaries of their training datasets, including origins and token counts.
The hosts stress this demands disclosure of trillions of tokens, copyright status, and exact data sources, which is far more granular than a typical summary.

INSIGHT

Disclosure Could Enable Competitor Reverse Engineering

Public dataset disclosures risk enabling reverse engineering of models by revealing data composition and filtering choices.
If regulators learn exact percentages of copyrighted text, synthetic blends, and conversational priorities, competitors get a huge head start replicating success.

Get the Snipd Podcast app to discover more snips from this episode

Get the app

Companies Complying with or Directly Impacted by Transparency Laws Major generative AI developers are broadly subject to AB 2013, which requires them to publicly disclose high-level summaries of the datasets used to train their models.

OpenAI, Anthropic, and Google were among the first companies to voluntarily comply with the law, publishing the required training data documentation on their websites when the law took effect on January 1, 2026.Meta is also heavily impacted by these laws and is frequently cited for its extensive efforts to harvest public and copyrighted data across the internet to train its foundation models.Companies Actively Challenging the LawxAI (founded by Elon Musk) is the primary company fighting the legislation. In late December 2025, xAI filed a federal lawsuit against California Attorney General Rob Bonta to block the enforcement of AB 2013. xAI argues that forcing it to disclose its training data constitutes an unconstitutional taking of its trade secrets and violates its First Amendment rights. In March 2026, a federal judge denied xAI's request for a preliminary injunction to halt the law.Separately, xAI is under investigation by the California Attorney General and received a cease-and-desist letter over its AI chatbot, Grok. The tool's "spicy mode" has allegedly been used to generate nonconsensual sexually explicit deepfakes and child sexual abuse material.Companies Sued Over AI Training Data and Copyright The push for transparency laws like AB 2013 and AB 412 stems largely from a massive wave of lawsuits filed by authors, artists, and media companies who allege that AI developers misappropriated their intellectual property to train models. Companies currently defending against these copyright lawsuits include:OpenAI and Microsoft (sued by The New York Times, The Daily News, the Authors Guild, Raw Story Media, and others).Anthropic (sued by Concord Music Group and various authors).Google and YouTube (sued by Mike Huckabee, David Milette, and others).Perplexity AI (sued by Dow Jones, The New York Times, and the Chicago Tribune).Stability AI, Midjourney, Runway AI, and Deviant Art (sued by visual artists and Getty Images).Meta, Nvidia, Databricks, and Mosaic ML.AI audio, music, and voice generation companies like Suno, Udio, Lovo, and ElevenLabs.Ross Intelligence (sued by Thomson Reuters for allegedly using copyrighted Westlaw data to train its own legal search tool).Other AI Companies Facing State ScrutinyCharacter.AI: Sued by the Kentucky Attorney General in January 2026 for consumer protection violations, alleging the company's companion chatbots preyed on children and contributed to psychological manipulation and self-harm. Google was also sued in related private litigation due to its substantial investment in Character.AI.Clearview AI: Cited by privacy advocates as a notorious example of unethical data sourcing, having scraped billions of images from social media to build a massive facial recognition database.