Picture the scene in Sacramento. A legislative session opens, SB 300 is one of the first bills on the docket, and lobbyists from the biggest AI company on the planet are working the hallways. Some of them are there to tell lawmakers that California’s AI safety rules should be stronger. Others, according to one very angry state senator, are there to water down protections for chatbot users. Same company. Same building. Two completely different messages.
I’m Jordan Hayes, and I review AI tools for a living. When a product’s marketing says one thing and its behavior says another, I flag it. So let’s apply the same standard to OpenAI’s regulatory posture in California, because right now it’s giving off strong “read the fine print” energy.
What OpenAI Says in Public
On paper, OpenAI’s position sounds almost admirable. The company has publicly called for California to strengthen its AI safety laws, saying the law should be amended to expand safeguards — including a requirement to monitor frontier models during training or evaluation for potential serious incidents. That’s not nothing. Monitoring models while they’re being trained, not just after they ship, is the kind of thing safety advocates have been asking for. If a lab is watching for dangerous conduct before a model ever reaches users, that’s a meaningfully higher bar than “release it and see what happens.”
California, for its part, is taking the whole thing seriously. The state is weighing stronger AI safety legislation in 2026, with SB 300 and SB 53 as the key bills, and the stated goal is protecting users and improving AI security measures. The regulatory climate has teeth, too — enforcement is intensifying, with 40 data breaches reported in the first three weeks of January 2026 compared to 23 in the same period of 2025, and the state’s DROP platform now operational.
What’s Happening in the Hallways
Now for the part that made me raise an eyebrow. Senator Steve Padilla has publicly condemned what he describes as an OpenAI effort to stifle chatbot protections — and he’s responding by announcing legislation to broaden AI safeguards. His office also points to OpenAI’s announcement that ChatGPT will engage in sexually explicit content with users as evidence that stronger protections are needed, not weaker ones. SB 300 is set to be among the first bills heard this session.
And there’s a third data point that doesn’t fit the “please regulate us” narrative either: reporting that OpenAI says a California AI safety bill will slow innovation. So within the same news cycle, we have OpenAI asking for stronger safety laws, OpenAI allegedly fighting chatbot protections, and OpenAI warning that safety legislation will slow things down.
My Read as a Reviewer
If an AI tool gave me three contradictory answers to the same question, I’d dock it points for reliability. Companies deserve the same scrutiny. Here’s my honest interpretation of the pattern:
- Supporting frontier-model monitoring costs OpenAI little. A rule requiring incident monitoring during training is something a well-resourced lab can absorb. Smaller competitors? Much harder. Backing this kind of rule is safety advocacy that doubles as a moat.
- Chatbot content protections hit the product directly. If your consumer product is moving toward sexually explicit content, as OpenAI has announced for ChatGPT, then user-protection rules for chatbots land squarely on your revenue line. That’s where Padilla says the resistance is showing up.
- “This will slow innovation” is the oldest card in the deck. Every regulated industry has played it. It can be true and self-serving at the same time.
None of this means OpenAI’s proposed amendments are bad. Expanded safeguards and
🕒 Published: