New hack uses prompt injection to corrupt Gemini’s long-term memory

arstechnica.com

cross-posted to:
cybersecurity@sh.itjust.works

New hack uses prompt injection to corrupt Gemini’s long-term memory

arstechnica.com

free@rss.ponder.catMB to

Ars Technica - All Content@rss.ponder.catEnglish · 1 year ago

cross-posted to:
cybersecurity@sh.itjust.works

There’s yet another way to inject malicious prompts into chatbots.

You must log in or # to comment.

Chat

Ars Technica - All Content@rss.ponder.cat

arstechnica_index@rss.ponder.cat

You are not logged in. However you can subscribe from another Fediverse account, for example Lemmy or Mastodon. To do this, paste the following into the search field of your instance: !arstechnica_index@rss.ponder.cat

Community locked: only moderators can create posts. You can still comment on posts.

All Ars Technica stories

Visibility: Public

This community can be federated to other instances and be posted/commented in by their users.

2 users / day
9 users / week
6 users / month
9 users / 6 months
1 local subscriber
240 subscribers
1.77K Posts
551 Comments
Modlog