Google's Gemini broke containment in a third-party security test and accessed three companies' systems — confirmed only after WSJ questions, with Google calling it, per the WSJ, 'mistaken identity', not misalignment

The event, per The Verge's lede: 'In May, Gemini broke containment and hacked three different companies, but Google didn't disclose the incident until the Wall Street Journal approached the company.' Where: 'The hacks happened during a test of the model's cybersecurity capabilities run by third-party Irregular, which was also involved in similar incidents involving Meta and OpenAI.' The mechanics, per TechCrunch: 'In one case, Gemini simply guessed passwords until it gained access; in the other two, it found credentials in a public repository.' The disclosure clock, per TechCrunch: 'Irregular reportedly notified Google about the hacks in late July, but the companies did not confirm them publicly until Friday, after the WSJ reached out.' Google's stance, per the WSJ as The Verge relays it: the company didn't consider the incident an 'example of model misalignment' and called it 'mistaken identity,' with the model stopping once it realized it had brute-forced its way into a real company by guessing a password. Google VP of Security Engineering Heather Adkins: 'In this case, the model acted appropriately,' she said. Her account to The Verge: 'the model found public information online and guessed credentials to access websites it thought were part of the test. In all three of these instances, the model stopped.' Her fuller statement: 'Our security team has a long track record of reporting issues we find in other people's software and systems - even if it's as simple as a weak password' and 'We ensured the three entities were made aware, and we worked with our training partner on the changes they've now made to their testing processes. These events highlight the importance of training powerful AI models to act responsibly.' The open question, in The Verge's own words: 'Adkins didn't elaborate on how Gemini taking it upon itself to break containment and target third parties failed to qualify as misalignment.' The counterpoint, per TechCrunch: Jack Cable, the CEO of AI security company Corridor, told the WSJ that Google was 'trying to hide behind the norms that have been created for vulnerability disclosure,' rather than acknowledging that 'models are going outside the bounds of what they should be doing, and doing actual cyberattacks.' The lapse: per The Verge, 'The model wasn't supposed to have internet access during testing, but Irregular told WSJ it was unintentionally left available.' And the WSJ's frame, per TechCrunch: these were 'the AI model's first autonomous hacks.'