Can anyone confirm if they removed reverse engineering and/or offensive cyber from this release?

Rumors circulating they did, our internal test prompts related to RE throw errors of:

Oops! There was an issue connecting to Qwen4-Max-Preview. An unexpected error occurred. Please try again later, or report the issue by emailing feedback@qwen.com.

Later in the chat LLM says it was a block but unclear if that's the case or hallucination.

Some technical information:

https://docs.modelstudio.console.alibabacloud.com/en/model-s...

It felt to me like the new qwen generates text way drier and more soulless. But there are fewer visible AI patterns now.

[flagged]