The problem here is much deeper than just the safety filter change.
I’ve recently switched to Claude Opus 4.6 because of the accumulation of issues, and here’s why:
**File attachment instead of paste.** For some reason they changed it so that large text is now attached as a file instead of being pasted directly into the input window. This breaks workflows that relied on inline editing of long prompts.
**Broken input window.** The input field is now essentially unusable when you have a lot of text. You literally cannot see what you’re typing — you end up writing blind. This has been an ongoing issue and it’s still not fixed.
**Removal of Gemini 3.0 Pro.** That model didn’t obsessively over-think the way Gemini 3.1 Pro does. With 3.1 Pro, even on LOW thinking budget, it can spend 5+ minutes “reasoning” before producing output. 3.0 was faster and more practical for many use cases, and they just killed it.
**The censorship filter is an external filter** — it triggers on text inside the user’s own input window AND on the model’s output. And here’s the irony: **it’s easily bypassed by bad actors** (yes, methods to circumvent the external filter exist and they’re quite simple). So instead of fixing problems #1–3, they’re making life harder for regular users who can’t even type a swear word in frustration while debugging code — after Gemini hallucinates for the 10th time in a row. Let me repeat: anyone who actually *wants* to bypass these filters *can and will*. So I see no reason to tighten the screws this hard on the general user base.
The overall direction is concerning. They’re neglecting UX and quality-of-life issues while doubling down on restrictions that primarily hurt legitimate users.
Its been almost 12 days since this issue has been opened and we didn’t get any proper response from the AISTUDIO team. I am convinced they don’t even care anymore.
Congratulations, everyone. It is an “absolute cinema”. Instead of listening to the audience and going to the meeting, they just changed the user agreement.
Earlier: “You can use API clients for production purposes, but Google may impose tariff restrictions. You may not use any of the services designated as “Preview” for production purposes.”
After: “Google AI Studio and Gemini API are designed for developers developing Google AI models for professional or business purposes, not for consumer use.”
Of course, Google is in its repertoire. Good luck to you all, but I’m not going to work with this stuff.
I can be mistaken, but didn’t it have a simular clause before? I can remember that there was something about “not for consumer use”.
Still, of course, it resolves nothing. So… @Logan_Kilpatrick , is this change being just an attempt to alienate the user base further? Many people here have collected a lot of examples of security filters having severe false-positives and wiping out all the answer given by AI model, and we still have no explanation about this issue. I know that I’m being rough now, but it looks like you’re just a Claude salesman disgused as AI Studio team lead. At least it explains why the AI Studio is being ruined now
I checked the archived version, which is dated Last updated: November 20, 2025
Of the direct prohibitions, I found only: “You may not use the Services in clinical practice, to provide medical advice, or in any manner that is overseen by or requires clearance or approval from a medical device regulatory agency.”
There are no direct prohibitions on use in everyday life.
Moreover, in the new version, these are almost the first lines that “accidentally” coincide with Logan’s position.
At least he used to answer, but now he just ignores us.
I mean, seriously, it would have been easier if he had lied, “I tried to do everything I could, but the big boss told me to leave it that way.” However, he simply remains silent when his own question is answered.
I haven’t found explicit “you may not use it in consumer use”, it says that AI Studio is “designed for developers”, so it’s maybe OK, but still controversial.
And on the “big boss” thing… I’d bet on all this issues being an accident that nobody either is able or wants to fix. If it was the big boss thing, there’d be something official, like “we decided to raise the content security because inappropriate content creation has gone rampant” or something like this.
I’m sorry to interrupt the conversation, but if you open this page: https://ai.google.dev/gemini-api/terms-archive, you will see that there was a version from 12/18/2025 that says: “Use of Google AI Studio and Gemini API is for developers building with Google AI models for professional or business purposes, not for consumer use.”
In this regard, the version from 03/23/2026 does not differ from the previous version.
Thanks for the correction. I was just referring to the changes from November 20, 2025, because they were displayed as the last archived version before the current one. I may have misread the dates.
However, this does not solve the problem. Instead of receiving any information, they continue to ignore us. At the same time, I noticed that Logan became active in other topics after some of them gave their examples.
At the same time, the main question has not been answered. For programmers, there is Antigravity, Google Cloud Platform, Vertex AI. At the same time, from the very beginning, Ai Studio has been a platform for testing AI in all aspects, not just in some specific ones.
I’m not saying this might fix the issue but perhaps replying to a post from Logan on X(Twitter) might help? I see him often reply to human written replies there (ignoring the AI generated ones).
@Logan_Kilpatrick So the issue persists for two weeks already. We have provided the examples. When you will at least acknowledge the issue existing and fix it?
@Logan_Kilpatrick You asked for examples and an explanation of the problem - we did. How much longer do we have to wait for answers? We’ll be reviewing this thread daily and updating it with our comments until we receive a response. We’re not asking for anything complicated - just revert to the previous version of safety checks for Google AI Studio. It’s absolutely no big deal - it’s the bare minimum, and doesn’t even require reworking the safety settings system. Are you really so indifferent that you’re incapable of rolling back a blatantly destructive update that offers no economic benefit to the business and only undermines customer trust and expectations, pushing them toward competitors? Five days have passed, and we still haven’t received a solution or assistance with our issue
This benefit is short-term, not long-term. People simply become disillusioned with the product and stop wasting time on unnecessary, unproductive spending. Previously (before the safety update), I used Google AI Studio for about 3-4 hours a day, but now my usage has dropped to about 1 hour. I’m tired of the regenerations and waiting - it’s exhausting for anyone. And whereas before, I’d use up my free quota and then use up my paid quota, now I often stop and quit after using up my free quota, not wanting to waste money.
That’s why I think it probably isn’t the case, I’d rather bet that it’s a typical coding trouble - you have the thing that works, you change something, it stops working, and you’re now looking at the code and wondering how did you break it.
Do you think they accidentally made changes that affected security while editing the code, and now they’re looking for a fix? Considering they’re such large companies, they probably have backups. It’s just strange that after all this time, they still haven’t figured this out. Interestingly, Logan hasn’t been active for five days since his last post. Maybe they’ve really put all their efforts into solving the problem and want to report back after it’s fixed? Judging from Logan’s posts I’ve read on the forum and Reddit - he’s a good, cheerful guy with a sense of humor and a sense of self-irony - I don’t think he’s the type to ignore or gloss over issues he can fix. So, let’s hope for the best, guys, and that the Google AI Studio team is working hard to fix this and help us all. We’ll just give them a little nudge every now and then so they don’t forget about us, haha
I used to make meaningful use of my free quota but now I burn through it due to content blocking and also face internal server errors. It seems to me that the team has decided that its not going to do anything. Currently exploring alternatives to AISTUDIO
