OpenAI and Microsoft back California's AI watermark bill

When you scroll past an ultra-realistic fake image on social media, you probably do not inspect the file's raw metadata to see if it was made by an algorithm. Neither does anyone else.

That gap between invisible digital breadcrumbs and what real people actually see on their screens is the core problem California lawmakers are trying to solve. Surprisingly, the industry's biggest players are now lined up to help them do it.

OpenAI, Adobe, and Microsoft sent letters supporting California bill AB 3211 as it heads toward a final vote in August, according to documents viewed by TechCrunch.

From hidden metadata to visible warnings

AB 3211 goes after synthetic media in two distinct ways.

First, it mandates that creators place watermarks directly inside the metadata of AI-generated photos, videos, and audio tracks. Many generative AI providers already do this voluntarily, but metadata remains completely invisible to the average user.

To close that loophole, the bill takes things a step further: it forces major platforms like Instagram and X to display clear, consumer-friendly labels whenever AI-generated media appears in users' feeds.

All three supporting companies are prominent members of the Coalition for Content Provenance and Authenticity (C2PA), which established the leading industry standard for embedding provenance metadata into media files.

A sudden corporate turnaround

The broad backing from Big Tech marks a dramatic shift in tone.

Just four months earlier in April, a major trade association representing Adobe, Microsoft, and other prominent software makers wrote a scathing letter to California lawmakers. At the time, the group slammed AB 3211 as "unworkable" and warned that it placed an "overly burdensome" compliance obligation on developers.

Subsequent amendments introduced to the bill addressed their primary objections, successfully flipping the industry's heaviest hitters from hostile opponents into active supporters.

Why it matters

For everyday internet users, this bill is a genuine win. It removes the guesswork from spot-the-fake games on your feed, shifting the burden of identification onto the platforms that profit from distributing the content.

For OpenAI, Microsoft, and Adobe, we think this pivot is a masterclass in strategic positioning.

Having already spent years building out the C2PA standard, they are essentially watching their preferred technical framework become the law of the land. Backing AB 3211 allows them to claim the high ground on AI safety while neatly shifting the heaviest technical and moderation burdens onto social networks like Instagram and X.

That said, clear limitations remain. Metadata is notoriously fragile; it often gets stripped during simple screenshots or aggressive compression on web platforms. Mandating labels is an important first step, but whether it can actually stop sophisticated deepfakes in the wild remains an open question.