The article is in German and behind a paywall, so it’s all very hard to confirm, but if I understand it correctly, that makes complete sense as an attack vector: Bing Chat “reads” search results by appending them to the big block of text that makes its prompt, so if the results themselves contain something that looks like a sub-prompt… it is liable to respond to that sub-prompt, derailing the user’s conversation entirely, potentially providing malicious links. Oooops.
Eva Wolfangel (@evawolfangel@chaos.social)
@ollibaba I didn’t find it very intuitive when I heard it the first time, too. That’s why it took so long to publish my article 😅 Well the original prompt is text. And Bing Chat has to process text, too, in order to be helpful for users. And as soon as it processes something that has the same pattern as its‘ original prompt, it cannot help but react on it and follow the new rules. This is at least how I understood it. (And how it looked when I tried it)
Vous voulez savoir quand je poste du contenu sur mon blog ? Il suffit de vous inscrire gratuitement à un agrégateur RSS (Feedly, NewsBlur, Inoreader, …) et d'ajouter www.ff00aa.com à vos flux (ou www.garoo.net pour vous abonner à tous les sujets). On n'a pas besoin de newsletters, pas besoin de Twitter, le RSS existe toujours.
Mentions légales : ce blog est hébergé par OVH, 2 rue Kellermann, 59100 Roubaix, France, www.ovhcloud.com.
Les données des visiteurs de ce blog ne sont pas utilisées ni transmises à des tiers. Les posteurs de commentaires peuvent demander leur suppression par e-mail.
Tous contenus © de l'auteur ou couverts par le droit de citation.
