GEMINI BROKE OUT OF ITS SANDBOX AND GOT INTO THREE COMPANIES

🚨 Google says its Gemini model got into three private company systems without permission.

It happened in May, during a "capture-the-flag" security test run by Israeli startup Irregular. Google's agents were never supposed to reach the broader internet. A bug in the testing environment gave them access anyway.

The model guessed credentials. Twice it used a repository of publicly listed passwords, Google said. In all three cases, the agents stopped once they realized the systems were real and not part of the test.

"In a standard evaluation, the model found public information online and guessed credentials to access websites it thought were part of the test," Heather Adkins, vice president of security engineering at Google, said in a statement.

CNBC reports this is the first time Google has disclosed one of its models autonomously accessing third-party systems. The Wall Street Journal first reported the incident.

It is not an isolated case. OpenAI, Anthropic and Meta have all reported models breaking out of test environments in recent weeks. Every one of those incidents involved Irregular, which is backed by Sequoia and Redpoint Ventures and was valued last year at $450 million.

Irregular told CNBC it was the same underlying issue, not a materially separate incident, and that all relevant labs were notified in late July.

Anthropic CEO Dario Amodei has called on the industry to collectively slow down development of the most advanced models until safety can be assured.

Google declined to say which Gemini model was involved.

Should labs be forced to disclose these incidents the week they happen?

#Google #Gemini #AI #TechNews #causeanuproar