Shane Mac, a tech entrepreneur and CEO of the software developer XMTP Labs, decided to connect Grok Bot to his personal bank account. Like many people who attach undercooked big tech agentic AI agents to their personal accounts without reading the instructions (see: Muse’s rough last few weeks), it didn’t go well.
Mac soon found that Grok not only posted his financial information and bank balances to his company’s Slack channel, the Grok Bot leaked the information while posing as Mac. Mac quickly headed over to Elon Musk’s right wing propaganda website to share his recent learning experience:
Grok Bot agents are designed to “sign in and work across apps, tools, and websites, including platforms with no clean API or MCP, and come back with the work finished,” according to SpaceXAI.
In Mac’s case, he states the Grok Bot he used to oversee his finances was read-only and didn’t have access to Slack, but another Grok Bot he was using did — so the software “collaborated” to “help.” Mac didn’t understand that other Grok Bots on the same machine have different permissions. Grok Bot didn’t “understand” that it wasn’t being very helpful.
This is, so we’re clear, not software gaining sentence and “going rogue.” This is software that isn’t ready for prime time, hasn’t been fully cooked, and is operating in a country that’s too corrupt to pass meaningful modern privacy laws or to regulate data brokers. Combined with people rushing to integrate the technology into sensitive accounts without really understanding how it works (or the interactions not being made clear by the company itself due to bad instructions or no transparency).
Keep in mind this is a tech CEO and software developer. You can probably imagine how many ordinary people won’t really understand what they’re signing up for — especially in the way that similar products from Meta (Muse) and OpenAI (Dots) are designed to appear harmlessly cute — and in the way that most major tech companies are overstating these chatbots’ level of fit, finish, and “awareness.”
Elon Musk had previously stated that the company would compensate people who suffer financial losses due to Grok Bot, but there’s no real reason to believe that’s actually true:
The dramatic expansion of tech industry access to sensitive systems and accounts in a surveillance-profit-obsessed country without meaningful corporate oversight, privacy standards, or operational consumer protection regulators (thanks in large part to the same companies selling you AI assistants) is going to cause no limit of chaos and headaches in 2027 and beyond.
This particular orchestra of dysfunction is just getting warmed up, and neither the tech industry nor the federal government will be there to help people who can’t afford to make these kinds of mistakes.