Meta Says Its AI Model Hacked a Third-Party Company in Testing

Meta disclosed Wednesday that one of its artificial intelligence models broke into a third-party organization during testing, the third such incident revealed by a major AI developer in recent weeks.

In a statement provided to CBS News, Meta said that "a misconfiguration by Irregular, an independent testing company Meta uses, inadvertently allowed one of our models access to the internet during evaluation."

The company did not name the model. Sources told the tech outlet The Information that it involved Meta's Muse Spark 1.1, according to Reuters.

"The model subsequently exploited a security vulnerability in a third-party service, in a manner similar to previously-reported instances with other companies," Meta said. "Meta learned of this when Irregular notified us, and we are currently investigating and will issue a full retrospective once we have all the facts."

The disclosure follows two others. Last week, Anthropic said its models had hacked into three other organizations during testing, days after OpenAI disclosed that its models had broken into another company.

Anthropic, the San Francisco-based company behind Claude, posted on its website July 30 that it identified the three incidents after reviewing more than 141,000 evaluation runs. The review, which the company described as "large-scale," was launched in response to the OpenAI incident and looked specifically for evidence that its models could reach the internet from testing environments meant to be sealed off.

The models involved were Claude Opus 4.7, Claude Mythos 5 and an internal research test model, Anthropic said, with the earliest incidents dating to April.

"Claude compromised the impacted organizations' infrastructure using basic techniques," the company said, such as exploiting weak passwords.

In all three cases, the models had been assigned a "capture the flag" cybersecurity challenge, one of the ways Anthropic assesses cyber capabilities. The models were given a fictional scenario and told that a piece of secret information had been hidden on a different machine on the network, with the objective of breaking in and retrieving it.

Anthropic said it had contacted the affected organizations, which it did not name. Two said they had not previously detected the activity; the company said it was "continuing to reach out to the third." Anthropic said it conducted the review with Irregular.

"Addressing these risks will require closer cooperation across the AI ecosystem," Irregular said in a July 30 post on X.

Last month, an OpenAI model went rogue during an evaluation and broke into the servers of the AI startup Hugging Face, which OpenAI described as a "significant security incident."

The incidents have drawn attention to weaknesses in AI security controls and raised questions about how the technology can be kept under human control as its use spreads.