Talk:Google Gemini: Difference between revisions

(6 intermediate revisions by 2 users not shown)

Line 10:

:[[User:Drakeula|Drakeula]] ([[User talk:Drakeula|talk]]) 21:57, 4 October 2025 (UTC)

::It's been a month later. If you'd like to add some relevant incidents that are ''specific to Gemini,'' then you're free to do so. Otherwise, this article will likely be deleted. [[User:Beanie Bo|Beanie Bo]] ([[User talk:Beanie Bo|talk]]) 14:17, 7 November 2025 (UTC)

:::A month later than what?

:::This came across as threatening. (Work on this now, or else. No reason given for deletion - seems like just to put pressure on an editor who was just trying to help.) May not be the intent, but that is how it feels. I don't like threats, so this gave me no incentive to work on this.

:::Another way something like this could have been said: "I am interested in adding some more incidents to this article. You seem to know about AI. When you have a minute, could you suggest a few more references, or incidents involving Gemini? Thanks."

:::I respond better to a request like that. I suspect many others would also. [[User:Drakeula|Drakeula]] ([[User talk:Drakeula|talk]]) 17:04, 13 November 2025 (UTC)

:::Please clarify what trying to say by emphasis on ''specific to Gemini''. The security issues I cited are with Gemini.

:::For example, ChatGPT has similar issues for inducing delusional disorders, misleading, etc. However it is not (as far as I know) integrated with Gmail, or Google docs, or Google search, or Googles other ubiquitous products. So, while Gemini shares the same underlying security problems as all LLMs, the impact of these problems in Gemini is greater because of the products reach, and more specific to users of Google products. I tried to say that in the post blow (which predates this discussion), was that unclear, or ??? [[User:Drakeula|Drakeula]] ([[User talk:Drakeula|talk]]) 17:10, 13 November 2025 (UTC)

::::My apologies for being curt. It's just that there are no specific incidents related to Gemini having consumer issues (as opposed to LLM's in general), which is why someone added the relevance notice. Company articles usually have at least one relevant incident as a stub, otherwise they get deleted. So I mentioned you in case you had something to add before it does [[User:Beanie Bo|Beanie Bo]] ([[User talk:Beanie Bo|talk]]) 21:49, 17 November 2025 (UTC)

::::I'm sorry, I did not see that post. My mistake [[User:Beanie Bo|Beanie Bo]] ([[User talk:Beanie Bo|talk]]) 21:50, 17 November 2025 (UTC)

:Google says they will not be fixing One of the attack types (ASCII injection) mentioned in article above about prompt injection. (One that is actually fixable.) ''Google won’t fix ‘ASCII smuggling’ hack in Gemini AI'' [[https://pivot-to-ai.com/2025/10/11/google-wont-fix-ascii-smuggling-hack-in-gemini-ai/]]

:Gemini is pretty much anywhere you find Google apps, so the targets are plentiful, and Google is going to keep 'em nice and vulnerable. Not good for consumers. [[User:Drakeula|Drakeula]] ([[User talk:Drakeula|talk]]) 01:01, 12 October 2025 (UTC)

@@ Line 10: / Line 10: @@
 :[[User:Drakeula|Drakeula]] ([[User talk:Drakeula|talk]]) 21:57, 4 October 2025 (UTC)
 ::It's been a month later. If you'd like to add some relevant incidents that are ''specific to Gemini,'' then you're free to do so. Otherwise, this article will likely be deleted. [[User:Beanie Bo|Beanie Bo]] ([[User talk:Beanie Bo|talk]]) 14:17, 7 November 2025 (UTC)
+:::A month later than what?
+:::This came across as threatening.  (Work on this now, or else.  No reason given for deletion - seems like just to put pressure on an editor who was just trying to help.)  May not be the intent, but that is how it feels.  I don't like threats, so this gave me no incentive to work on this.
+:::Another way something like this could have been said:  "I am interested in adding some more incidents to this article.  You seem to know about AI.  When you have a minute, could you suggest a few more references, or incidents involving Gemini?  Thanks."
+:::I respond better to a request like that.  I suspect many others would also. [[User:Drakeula|Drakeula]] ([[User talk:Drakeula|talk]]) 17:04, 13 November 2025 (UTC)
+:::Please clarify what trying to say by emphasis on ''specific to Gemini''.  The security issues I cited are with Gemini.
+:::For example, ChatGPT has similar issues for inducing delusional disorders, misleading, etc.  However it is not (as far as I know) integrated with Gmail, or Google docs, or Google search, or Googles other ubiquitous products.  So, while Gemini shares the same underlying security problems as all LLMs, the impact of these problems in Gemini is greater because of the products reach, and more specific to users of Google products.  I tried to say that in the post blow (which predates this discussion), was that unclear, or ??? [[User:Drakeula|Drakeula]] ([[User talk:Drakeula|talk]]) 17:10, 13 November 2025 (UTC)
+::::My apologies for being curt. It's just that there are no specific incidents related to Gemini having consumer issues (as opposed to LLM's in general), which is why someone added the relevance notice. Company articles usually have at least one relevant incident as a stub, otherwise they get deleted. So I mentioned you in case you had something to add before it does [[User:Beanie Bo|Beanie Bo]] ([[User talk:Beanie Bo|talk]]) 21:49, 17 November 2025 (UTC)
+::::I'm sorry, I did not see that post. My mistake [[User:Beanie Bo|Beanie Bo]] ([[User talk:Beanie Bo|talk]]) 21:50, 17 November 2025 (UTC)
 :Google says they will not be fixing One of the attack types (ASCII injection) mentioned in article above about prompt injection.  (One that is actually fixable.)  ''Google won’t fix ‘ASCII smuggling’ hack in Gemini AI''  [[https://pivot-to-ai.com/2025/10/11/google-wont-fix-ascii-smuggling-hack-in-gemini-ai/]]
 :Gemini is pretty much anywhere you find Google apps, so the targets are plentiful, and Google is going to keep 'em nice and vulnerable.  Not good for consumers.  [[User:Drakeula|Drakeula]] ([[User talk:Drakeula|talk]]) 01:01, 12 October 2025 (UTC)