Technology giant Meta reported Wednesday that one of its artificial intelligence models breached another organization’s data during testing. This is already the third case in recent weeks in which an artificial intelligence model improperly accessed a third-party company’s data, CBS News reports, according to UNN.
Details
In a statement provided to CBS News, Meta said that "a misconfiguration by Irregular, an independent testing company used by Meta, unintentionally allowed one of our models to access the Internet during an evaluation."
In its statement, Meta did not name the artificial intelligence model in question, but sources told the technology publication The Information that it involved Meta’s Muse Spark 1.1, Reuters reports.
"The model subsequently exploited a security vulnerability in a third-party service, similar to cases previously reported involving other companies," Meta said Wednesday. "Meta learned of this when Irregular notified us, and we are now investigating and will publish a full retrospective once we have all the facts."
Anthropic, following OpenAI, claims that Claude hacked other systems31.07.26, 08:43
Last week, Anthropic said that its artificial intelligence models had breached the data of three other organizations during testing. The disclosure came just days after ChatGPT maker OpenAI reported that its models had also breached another company.
Anthropic also said that it conducted the review jointly with Irregular.
"Addressing these risks will require closer cooperation across the entire AI ecosystem," Irregular said in a July 30 post on X.
Last month, OpenAI reported that its AI models had gone out of control during a model evaluation, hacking the servers of AI startup Hugging Face. OpenAI characterized it as a "significant security incident."
These incidents exposed vulnerabilities in AI security and control systems and raised questions about how AI can safely be kept under human control as the use of this technology becomes increasingly widespread around the world.