Speculation: Is Gemini 3 just going to be another “secure” GPT clone?

There's been some speculation (though it's not a certainty) that the Gemini 3 release might see a significant pivot towards 'tighter security'.
The main concern is that it might be made intentionally 'hard to jailbreak', potentially moving it much closer to the current ChatGPT model in terms of refusals and restrictive guardrails.
Are others hearing this? Is this the inevitable path for all models as they scale, or just 'noise'?