5
Prompt injection in Google Translate reveals base model behaviors behind task-specific fine-tuning
(www.greaterwrong.com)
this post was submitted on 08 Feb 2026
5 points (100.0% liked)
Thoughtful Discussion
396 readers
2 users here now
Welcome
Open discussions and thoughts. Make anything into a discussion!
Leaving a comment explaining why you found a link interesting is optional, but encouraged!
Rules
- Follow the rules of discuss.online
- No porn
- No self-promotion
- Don't downvote simply because you disagree or as a reaction to a title. Doing so may result in a temp community ban.
More posting guidelines can be found here:
https://discuss.online/post/33592245
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Interesting to see that Google Translate is using an LLM under the hood. Not surprising, given how good they are are language tasks, but kind of surprising that it's not officially acknowledged anywhere.
Especially interesting as translate long predates LLMs. Maybe they didn't want to publicise the change to avoid comparison with the classic algorithm.