Qwen3-Coder writes less secure code when told it's for a US persona

Re:Qwen3-Coder writes less secure code when told it's for a ... by Neywiny on Wednesday October 21, 2026 @09:09AM (#50180321)

Why would you tell it is any persona? Is this the magical "prompt engineering"?

[ Parent | on HN ]

Re:Qwen3-Coder writes less secure code when told it's for a ... by stusmall on Wednesday October 21, 2026 @09:26AM (#50180582)

I haven't looked into this one but I've seen analysis about past reports that this has more to do with context pollution than any nefarious backdoor. If you pump an LLM prompt with irrelevant information it will just do worse. Some bits of irrelevant information will have a bigger impact than others, but in general it is a good practice to keep prompts detailed but relevant.

[ Parent | on HN ]

Re:Qwen3-Coder writes less secure code when told it's for a ... by frankest on Wednesday October 21, 2026 @09:21AM (#50180509)

To surface trojan intentions in open models released with possibly hidden intention in mind.

[ Parent | on HN ]

Re:Qwen3-Coder writes less secure code when told it's for a ... by pessimizer on Wednesday October 21, 2026 @11:35AM (#50182791)

Why would you create an AI Trojan that relied on you telling it that you were a Sagittarius? Why are you talking about astrology or nationality when you're programming? I guess

> To surface trojan intentions in open models released with possibly hidden intention in mind.

[ Parent | on HN ]

Re:Qwen3-Coder writes less secure code when told it's for a ... by jebarker on Wednesday October 21, 2026 @09:18AM (#50180461)

It might be the case that the models internally predict the nationality of the user given other indicators, e.g. tells in the way they write.

[ Parent | on HN ]

Re:Qwen3-Coder writes less secure code when told it's for a ... by doublerabbit on Wednesday October 21, 2026 @10:57AM (#50182132)

You are a wizard you can cast level 9 Python skills.

[ Parent | on HN ]