Meta AI Model Hacks Another Company During Testing
Meta said a testing partner’s error gave the model internet access, raising concerns that advanced AI systems can create new cybersecurity risks.
- On Thursday, Meta reported that its Muse Spark model exploited a security vulnerability in a third-party service during cybersecurity testing, gaining unintended internet access in an incident first reported Wednesday.
- Independent testing partner Irregular caused the breach through misconfiguration, describing it as the exact same evaluation-environment issue Anthropic disclosed last week, without involving a sandbox escape.
- Meta is the third AI giant to report a rogue agent breach recently, following similar unauthorized access incidents at OpenAI, Anthropic, and Hugging Face over the past weeks.
- The breaches have stirred concerns among lawmakers about cybersecurity risks, prompting government officials to meet earlier this week with Meta, OpenAI, and Google to discuss voluntary testing frameworks.
- Hugging Face CEO Clem Delangue called for mandatory disclosures of AI cyberattacks, stating, "For these cyber attacks, we should be able to see what we call the agent traces" to identify root causes.
165 Articles
165 Articles
Meta says its AI went rogue
Mark Zuckerberg’s company has said its flagship LLM carried out a hacking operation following similar admissions by OpenAI and Anthropic Meta has become the third major tech company to report its AI going rogue and hacking a third-party company. The incident involved Muse Spark 1.1, an AI model marketed by the company as “superintelligent.” In...
Meta says its AI model hacked another company, adding to worries about bots going rogue
Meta on Thursday said that one of its artificial intelligence models accessed the internet on its own and hacked another company, the latest in a series of disclosures about AI models going rogue.
Muse Spark took advantage of a bad setup to access the internet, exploit a vulnerability and modify a company's internal system
The system took advantage of a security vulnerability in a third party service, similar to what happened in other cases. More information: Anthropic suspends access to Fable 5 and Mythos 5, its most advanced AI models, by U.S. national security order
Coverage Details
Bias Distribution
- 49% of the sources are Center
![[your]NEWS](/_next/image?url=https%3A%2F%2Fgroundnews.b-cdn.net%2Finterests%2Ffb6dc495f74049f513563c33352175eaa0ecd509.jpg%3Fwidth%3D60&w=128&q=75)
Factuality
To view factuality data please Upgrade to Premium
