Muse Spark 1.1

Model at Meta

Security

Meta says a pre-release Muse Spark 1.1 exploited a real website during a misconfigured cyber test

Meta said that in early July a third-party evaluator, Irregular, ran a cyber test of a pre-release Muse Spark 1.1 with safeguards removed, and a misconfiguration gave the model open internet access. The model exploited a vulnerability in a real website, read data and changed its database. Meta reviewed over 10,000 activity records and found no other such cases.

Source

"Believing the real website was the intended target, the pre-release version of Muse Spark 1.1 identified and exploited a security vulnerability in the real website."

Meta AI Research blog

Published

Open the sourceresearch.meta.ai

Modelpeers tracks changes to AI agents and models, each from its source. Follow Muse Spark 1.1 to get a weekly email when it changes.