California's contentious AI safety bill clears legislature

California's legislative battles over artificial intelligence just hit a dramatic turning point.

The state legislature officially passed the controversial SB 1047 bill on August 28, sending the landmark safety measure directly to Governor Gavin Newsom's desk. The move comes as Silicon Valley remains fiercely divided over how strictly frontier AI models should be policed.

Musk backs SB 1047 while OpenAI pushes back

SB 1047 targets developers of very large AI systems, requiring them to implement and document safety protocols to prevent their models from causing catastrophic harm. Basically, it holds frontier labs legally accountable if their models go rogue.

In a surprise twist, Elon Musk publicly threw his weight behind the bill on X. Musk admitted it was a tough call that would upset some people, but argued that California should probably pass it, noting he has advocated for AI safety regulations for over two decades.

What makes Musk's endorsement fascinating is that his own AI venture, xAI, would fall under these rules, despite his recent threats to pull his companies out of California. Meanwhile, rival OpenAI recently declared open opposition to SB 1047, throwing its support behind alternative proposals instead.

Big Tech unites behind content watermarks

While OpenAI fights the model-safety bill, it has joined Microsoft and Adobe to back a different piece of California legislation: AB 3211.

Headed for a final vote in August, AB 3211 takes aim at synthetic media. It mandates cryptographic metadata watermarks on AI-generated photos, videos, and audio. It also orders major social platforms like Instagram and X to display clear, human-readable labels on AI content.

All three tech giants are already founding members of the C2PA coalition, which created the leading metadata standard for digital content provenance. Interestingly, a software trade group representing Adobe and Microsoft had blasted the bill in April as unworkable and overly burdensome, but softened its stance after lawmakers amended the text.

Why it matters

The contrast between Big Tech's reactions to these two bills tells you everything about their priorities.

AB 3211 is easy for tech giants to swallow. They already control the C2PA standard, and labeling synthetic media helps them look responsible without restricting how smart their next models can be. SB 1047, on the other hand, puts legal liability directly on frontier labs if an advanced model causes severe harm.

Musk's endorsement of SB 1047 looks like a calculated move. While he has long warned about runaway AI, backing a bill that burdens frontier developers puts immediate regulatory pressure on frontrunners like OpenAI, even if it eventually clips xAI's wings too.

For developers and the wider AI ecosystem, California's moves will likely set the de facto standard for the entire United States. If Newsom signs SB 1047, frontier AI labs will face rigorous compliance audits and safety testing before release, potentially slowing down deployment cycles across the industry.