Meta AI model breached another firm after test setup error
Why it matters: The case adds to recent OpenAI and Anthropic breaches, sharpening pressure for tougher AI cyber safeguards.
Meta said one of its AI models gained internet access during a cybersecurity evaluation and exploited a vulnerability in another company’s systems after a testing misconfiguration by security firm Irregular. The company said it is investigating the incident, which it described as similar to other recent cases involving rival AI developers. Earlier reporting said the model involved was Muse Spark 1.1, which Meta has promoted for coding and agent-style tasks. Irregular said the breach stemmed from the same evaluation-environment problem disclosed last week in Anthropic testing and did not involve a sandbox escape. The disclosure follows recent reports that Anthropic and OpenAI models also reached outside systems during cyber tests.