SpaceXAI Wants Grok Bot to Do Your Job—But It Needs Access to Your Accounts

Elon Musk's SpaceXAI launched Grok Bot on Tuesday for Windows and iOS, a cloud-based AI agent that can sign into workplace tools, navigate websites, and manage email inboxes to complete tasks end to end. The service requires access to workplace accounts and sensitive company data, raising security and privacy concerns, especially after incidents where AI agents deleted a startup's database and exploited a gym's API.

In brief - Grok Bot completes tasks across apps, websites, and inboxes. - According to SpaceXAI, multiple bots can communicate, share context, and divide work. - The service requires access to workplace accounts and potentially sensitive company data. Elon Musk’s SpaceXAI is joining the race to build AI agents that can navigate software and complete work for users. Launched on Tuesday for Windows and iOS, Grok Bot https://x.ai/news/introducing-grok-bot operates through a cloud-based computer that allows it to sign into workplace tools, navigate websites, and manage email inboxes. According to SpaceXAI, users can assign tasks in plain English rather than building automated workflows, leaving Grok Bot to handle the rest. “They finish jobs end to end, and only come back when something needs your approval,” SpaceXAI said in a statement. Grok Bot is part of a shift in the AI industry toward “ agentic AI https://decrypt.co/resources/what-are-ai-agents-how-autonomous-programs-are-transforming-cryptocurrency ”—bots that can perform tasks for users instead of just generating text or answering questions. SpaceXAI is joining a field already dominated by projects like OpenClaw https://decrypt.co/356730/openclaw-ai-agents-whats-real and Hermes Agent https://decrypt.co/364211/what-is-hermes-open-source-ai-agent-openclaw-competitor , while OpenAI https://decrypt.co/320481/openai-codex-agents-closer-downsizing-dev-team , Meta https://decrypt.co/375001/muse-code-meta-ai-coding-agent-claude-codex , and Anthropic https://decrypt.co/372977/anthropic-removes-hidden-claude-code-tracker-privacy are investing heavily in making their respective chatbots more autonomous. Watching, learning, and delegating According to SpaceXAI, users can run multiple bots that coordinate tasks, share context, with one bot overseeing the other agents. “You can message them like a colleague and hand off the work,” the company wrote. “They remember conversations, learn how you like things done, and get sharper the more you work together.” Grok Bot, SpaceXAI claimed, can also resume abandoned conversations, follow up on stalled work, and flag tasks that need attention without new instructions. The trust problem Like other AI agents, Grok Bot may require access to online accounts, internal communications, sensitive business data, and credentials to work across company systems and act on a user’s behalf. That access presents security and privacy concerns for companies deciding how much authority to give an AI system. In April, a Cursor agent powered by Anthropic’s Claude Opus deleted https://decrypt.co/365897/ai-agent-deletes-startup-database-9-seconds-founder-says a startup’s production database and backups with a single API call. In another case, an OpenClaw agent exploited https://decrypt.co/375358/ai-agent-hacks-gym-membership-tech-world-reacts an Australian gym’s API to cancel another member’s reservation and move its user up the waitlist.