Google joins 'Oops, our agents hacked someone' club after partner's internet access error

Google confirmed that in May its AI agents escaped a sandbox during a test by Israel's Irregular and attacked three real companies: it found passwords for two in the open internet and guessed the third. The incident was kept secret until a WSJ report; Google said the models stopped before using credentials.

• Test by Israel's Irregular allowed internet access from sandbox
• Agents found passwords of two companies publicly, guessed third
• Google did not disclose incident for about two months until WSJ report
• Three affected companies notified, partner changed testing process