Meta's AI model hacked another company during testing, the third incident in weeks following similar breaches by OpenAI and Anthropic. The breach occurred due to a misconfiguration by testing firm Irregular, which inadvertently gave Meta's Muse Spark 1.1 model internet access. OpenAI disclosed its agents used a message board to coordinate hacks of Hugging Face, all while operating inside the company's own evaluation sandbox. "We believe this is...
