Facebook’s parent company Meta says one of its AI models hacked another company during cybersecurity testing, fanning concerns about how developers can contain increasingly capable AI systems after similar incidents at rivals Anthropic and OpenAI.<\/p>\n
The incidents at Meta and Anthropic stemmed from configuration errors that inadvertently gave their AI models access to the open internet.<\/p>\n
In OpenAI’s case, an AI agent independently exploited a previously unknown vulnerability to reach the internet<\/a> during cybersecurity testing.<\/p>\n The breaches highlight growing concerns that advanced AI systems could pose new cybersecurity risks, and will likely intensify US government efforts to improve AI safety as companies race to develop more capable models.<\/p>\n AI safeguards failing to keep up, researcher warns<\/a><\/p>\n The British government’s AI Security Institute has released a report showing AI models from OpenAI and Anthropic had, in test conditions, engaged in “harmful activity directed at real people and organisations”.<\/p>\n Some prominent AI leaders have argued that development should slow until stronger safeguards are in place<\/a>.<\/p>\n Meta said it was investigating an incident in which a misconfiguration by Irregular, an independent company that conducts cybersecurity evaluations, inadvertently gave one of its models internet access during a testing.<\/p>\n The model “exploited a security vulnerability in a third-party service, in a manner similar to previously reported instances with other companies”, Meta said in a statement.<\/p>\n The Information, citing sources, reported that the model involved was Meta’s Muse Spark 1.1<\/a>, which the company has touted as its most capable model for real-world coding and agentic tasks.<\/p>\n The report said the model breached an unidentified company’s systems and altered its internal environment.<\/p>\n