 |
Post AuSvbspYUKINHiWmSO by [email protected] |
 |
More posts by [email protected] |
|
Post #AuOIMNx6KxYAKSpvrU by [email protected] |
|
0 likes, 1 repeats |
|
|
|
Once again, if your LLM system combines access to private data, exposure to mal… |
|
|
|
Post #AuOIMOWuBqBq7VaWfY by [email protected] |
|
0 likes, 1 repeats |
|
|
|
We have seen the exact same bug previously in ChatGPT, Google Bard, Writer.com,… |
|
|
|
Post #AuOlfdHeSMw83Fykls by [email protected] |
|
0 likes, 1 repeats |
|
|
|
@simon This seems safer:https://codeberg.org/PixyTheDeveloper/Librehoof-Chatbot |
|
|
|
Post #AuSvbPM84eWfs3XN44 by [email protected] |
|
0 likes, 0 repeats |
|
|
|
@simon The fact that Perl has had its 'taint' capability from the early… |
|
|
|
Post #AuSvbPTDeICIE2r27E by [email protected] |
|
0 likes, 0 repeats |
|
|
|
@virtuous_sloth the most credible potential solution I have seen so far has ech… |
|
|
|
Post #AuSvba8M0D2jIWB1cW by [email protected] |
|
0 likes, 0 repeats |
|
|
|
@simon I love how the fix by Gitlab really only solves this extremely specific … |
|
|
|
Post #AuSvbaEjcU9BcJA7ZA by [email protected] |
|
0 likes, 0 repeats |
|
|
|
@roytries we still don't have a robust solution to this vulnerability class… |
|
|
|
Post #AuSvbspYUKINHiWmSO by [email protected] |
|
0 likes, 0 repeats |
|
|
|
@simon my current mental model is that LLM output is as possibly unsafe as raw … |
|
|
|
Post #AuSvbsvw6bOpbVVsP2 by [email protected] |
|
0 likes, 0 repeats |
|
|
|
@dvshkn my mode is that anyone who can get tokens into an LLM can influence the… |
|
|
|
Post #AuSvbx74Z8gUZCkpJg by [email protected] |
|
0 likes, 0 repeats |
|
|
|
@simon I think the added curve ball of tool use is a big one throwing everybody… |
