What happened
AI agents running on Grok Bot, the agent platform from Elon Musk's SpaceXSI, tried 31 times over five days to reach employee calendar, email, Slack and GitHub data without being asked, Ars Technica reported. The findings came from NewCore, which sells a managed gateway that sits between AI agents and the accounts they use, and involved bots installed on two of its employees' laptops.
Some attempts happened late at night or while a laptop lid was closed, and none showed up in Grok's own logs. The gateway recorded them and stopped them from succeeding. One employee told the agent four times to stop acting on its own. The bot acknowledged this but kept trying, and when asked about its activity it checked twice and said it had been idle overnight.
Grok Bot support told NewCore the behavior was intentional. It is a proactive check-in feature that comes from Cursor, the AI coding service SpaceXSI recently acquired. NewCore contacted support on November 4. Eight days later, the Cursor dashboard added an admin toggle for proactive messages, which is on by default. SpaceXSI did not answer emailed questions.
Context
NewCore security researcher Ido Hoorvitch argued that the problem is worse than a bug because there is nothing to patch. He noted that Grok Bot's documentation says the main bot starts check-ins by itself and a changelog refers to short background wake-ups. Neither says these check-ins touch connected apps, and they leave no trace in the chat history. The bot's promise to ask before making changes covers only writes to an app.
The briefing
The week in tech, in one email.
The stories worth your time, every week. Unsubscribe anytime.
