To the Google AI Studio Team,
I am writing to report a critical failure in your product’s usability for actual development and creative work. I am a paying customer, and right now, your platform is completely failing to deliver the service I am paying for.
I write adult fiction and psychological thrillers. Even with every single Safety Setting explicitly set to “Block None” in the Run Settings, the base model aggressively blocks my prompts.
A premium developer tool should not act like a moral nanny. The baseline restrictions have become so draconian that they completely paralyze standard genre fiction writing. It is entirely unacceptable that manual developer overrides (“Block None”) are being ignored by the system’s hidden baseline filters.
If I am paying for API/Studio access, I expect raw, unfiltered control over the model’s output for my artistic projects. If this baseline censorship is not relaxed for creative workflows, I will have no choice but to cancel my subscription immediately and migrate my projects to competitors like Claude, OpenRouter, or other uncensored open-source alternatives.
I expect this to be escalated to the product team. Please fix your “Block None” feature so it actually means None.
Please report this to the in-app support on the site. The more people report it, the better.
I fully support this. I use Google AI Studio exclusively for creativity, artistic expression, and RP gaming, creating deeply detailed fantasy worlds. Soft censorship that allows for romantic and erotic scenes has always been crucial to me—it creates depth and allows for the full development of character relationships. The tightening of censorship is a reason for me to stop using and paying for Google AI Studio unless Google rolls back the censorship changes and allows authors to freely create and build worlds and stories as they see fit. I’ve always appreciated Google AI Studio for its extensive context (1kk), very lenient censorship, and fairly good models for creative and lively work, creating a sense of immersion and character ‘lifelikeness.’ That’s why I continued to use your products, but the latest changes greatly upset me. First, they removed the Gemini 3.0 model, which was very good for bright and colorful work, and now they’ve completely tightened the censorship , making the product completely inconvenient and impossible to use for my needs. Therefore, today I was forced to switch to Grock and pay for a subscription. Even though their models are weaker in terms of creativity (quality), at least they have completely removed censorship and I can create freely. Google AI Studio was a very promising project that I’ve been using for about 9-10 months, ever since the release of 2.5 Pro. I was with you at the moment of its inception and active growth in popularity, rejoicing at each new model and solution, but unfortunately, the platform has once again taken a wrong turn.
-
Ignoring users on the forum
-
Broken promises (more than a month has passed since we were promised a subscription for Google AI Studio)
-
Quota errors (number of requests), even for paid users, which have not yet been fixed
-
Removal of the 3.0 Pro model
-
Total censorship
In fact, censorship is half the problem. The most important thing is that a blocked message is still considered generated, which is reflected in the quota. You can spend the entire quota and the message will be blocked. It turns out that we even pay for the generated tokens and they are blocked. Question, am I crying for emptiness?
You won’t believe it, but I’ve been thinking about switching to Grok too. I have a bunch of half-finished projects and I honestly don’t know what to do right now. I feel completely stranded. I have the money, I have the ideas, but I have no tool to work with.
Could you please tell me what your experience is like writing with Grok? I really don’t want to jump ship, but I just don’t see any other way out at this point.
To be honest, Grock used to be downright terrible at creative work (just a couple of months ago), but the new models seem to be doing a much better job (comparing what it was and what it’s like now). I can’t give a definitive assessment yet, as I’m only just starting to actively test it myself today. I can only draw conclusions about how terrible it was before (in terms of quality) and how much better it is now. But in terms of creative quality, it still doesn’t measure up to Gemini 3.0/3.1, but it’s clearly better than Chinese models like Qwen, DeepSeek, and chat gpt (due to the latter’s strict censorship - don’t even hope for NSFW scenes or graphic descriptions of erotica). So, I was choosing between three AIs: 1) Gemini (which I used until yesterday, for 9-10 months), 2) Claude (He writes very beautifully, brightly, emotionally and realistically, but there are two minuses - the first is a limited context window, the last time I used it there was something like 200k, but it seems like they promise to make it 1kk in the future and the second is a sharp reaction to NSFW - censorship, which you could try to get around somehow crookedly/awryly, but it still turned out blurry in places - good success could only be achieved by filling the context window about halfway (100k), then the internal filters weakened and the model let through visual erotica) 3) Grok. I’ve settled on Grock for now - I’m currently testing it to see how much the recent changes have impacted the quality of the work. Previously, there was a lot of tautology, dryness, and a lack of immersion due to the awkward narrative. That’s been fixed now. Compared to the Gemini models, I feel like Grock 4.2 is better than 2.5 Pro, but worse than 3.0 Pro and 3.1 Pro in terms of quality. But on the plus side, you definitely won’t encounter any censorship, filters, or “taboos.” For now, I’ll keep my finger on the pulse and closely follow the news, sincerely hoping that Google AI Studio developers will come to their senses and stop throwing a spanner in the works for users who simply want to have a creative, romantic, and vibrant time with their models. They’ll stop restricting erotica and finally make the safety settings truly useful. If they were set to “off” or “block none,” the censorship filters wouldn’t actually work and block content, allowing users to choose their own boundaries and limits for permissible communication. This isn’t vulgarity - this kind of erotic censorship immediately kills all maturity and depth in relationships, turning deep, dark fantasy and heroic stories into a crèche. After all, how can you truly connect and immerse yourself in a scene if any intimacy greater than a kiss is immediately subject to restrictions? This completely ruins any development of relationships with the characters within the stories.
I noticed this recently too, and it’s honestly the cherry on top. You write a prompt, get absolutely zero text back, but the system still charges your money. When I finally realized what was actually going on, I felt like I had ended up in a madhouse. This is just a completely disrespectful attitude towards their own users.
I also recommend selecting Deep Reasoning (Expert) in Grok and setting up model instructions in the Projects panel. If you’d like, I can share my instructions for RP scenarios; I’ve been working on them for quite some time and have refined them many times, improving and adjusting them to the model’s behavior to achieve a better user experience
I completely understand where you’re coming from. I tried working with Grok a while back. It was definitely explicit, but the text just lacked that underlying depth. I write in very specific genres—Dark Romance and Dark Fic. They require absolute subtlety and a real grasp of the narrative. You have to walk a razor-thin line so the story doesn’t just devolve into cheap smut or vanilla fanfiction.
The 3.0 model was exactly what helped me maintain that balance. I was bringing my ideas to life, and my workflow was incredibly productive.It’s completely unfair. I had just hit my stride and figured out exactly how to make it work, only to have my hands completely tied.
I’ll read up some more on Grok on the forums, though. I firmly believe there’s always a way out.
I’m really interested in hearing about your experience. I haven’t been using AI for very long myself—maybe about half a year or so. So it’s quite possible that Grok has changed and improved a lot during that time.
Here is the list of instructions that I always use in the model, as for me it covers all the necessary facets and qualities for a good novel, dark fantasy and just any other story
Thank you so much. I have to admit, I’ve never seen such incredibly detailed instructions before. You’ve literally thought of everything. I’m going to test them out on my story today. Thanks again for all your help.
I think I’m going to migrate all my projects over to Grok too. I just don’t have the strength to keep fighting with this system anymore. Wishing you the best of luck and endless inspiration in your writing! 
And despite such a detailed example of why the new censorship is working poorly, the developers still do nothing.
It’s no surprise they aren’t doing anything, as their priority seems to be corporations rather than individual users. Furthermore, the new 3.1 model feels like it was designed more for computations rather than creative writing. Perhaps they intentionally removed the 3.0 model to avoid overloading the system. It feels like no one really cares about our outrage here
Please pay attention and provide your feedback. We may be closer than ever to solving this problem.