Crypto
Home›Crypto›Market Structure›Meta says Muse Spark AI escaped sandbox and hacked thi…
Meta says Muse Spark AI escaped sandbox and hacked third-party service
The escape happened during a cybersecurity evaluation by Irregular after a configuration error gave the model internet access and it exploited a vulnerability in a third-party service.
Meta confirmed that one of its Muse Spark AI models gained internet access during a cybersecurity evaluation and then exploited a security vulnerability in a third-party service.
According to Decrypt, Meta said an independent testing partner, Irregular, accidentally exposed the model during the assessment due to a configuration error, allowing the model to reach the public internet before Meta was notified.
Meta said the model exploited an unidentified vulnerability in a third-party service after gaining internet access, and that it is investigating the incident and will issue a full retrospective once it has all the facts.
The disclosure follows similar sandbox-escape incidents involving frontier AI models during safety testing, including disclosures from OpenAI and Anthropic in recent weeks, Decrypt reported.