Blog

New attack on ChatGPT research agent pilfers secrets from Gmail inboxes

Unlike most prompt injections, ShadowLeak executes on OpenAI’s cloud-based infrastructure.  The face-palm-worthy prompt injections against AI assistants continue. Today’s installment hits OpenAI’s Deep Research agent. Researchers recently devised an attack that plucked confidential information out of a user’s Gmail inbox and sent it to an attacker-controlled web server, with…

Some dogs can classify their toys by function

Come for the science. Stay for the adorable pics of happy pupsters with their favorite toys. Certain dogs can not only memorize the names of objects like their favorite toys, but they can also extend those labels to entirely new objects with a similar function, regardless of whether or…