Malicious LLMs Are A Thing
“Mallas” may sound like innocent shorthand for Kamala Harris supporters or marshmallow lovers, but it actually describes malicious LLMs (large language models). Academics at Indiana University Bloomington found 212 such systems on underground marketplaces, using a mix of open source and commercial platforms. If there’s any good news, it’s that Malla-created phishing content was easier to spot than human-created phishing content.