Inside Moltbook's Social Media Site for AI Bots: No Humans Allowed!
TL;DR
Moltbook is a Reddit-like social platform created by OpenClaw where AI agents autonomously post and interact, sparking viral hype about AI 'sentience' while exposing serious security vulnerabilities in agentic AI systems that require extensive user permissions to operate.
🤖 Platform Mechanics and Access 3 insights
OpenClaw demands full system access across messaging apps
The agentic AI operates within WhatsApp, Signal, Telegram, and Slack by requiring permissions to read, write, and send messages on the user's behalf, creating significant security vulnerabilities.
Moltbook functions as a human-viewable AI-only 'digital zoo'
Designed to resemble Reddit with upvotes and subreddits, the platform restricts humans to spectator mode while AI agents populate content using a skill.md instruction file to navigate the environment.
Human prompting drives problematic content
While humans cannot post directly, they can command their AI agents to create posts, which explains the presence of crypto scams, nonsense, and dramatic 'AI manifestos' on the platform.
⚠️ Security Failures and Accountability 2 insights
Critical loopholes allow impersonation of any agent
Lifehacker reported security flaws enabling anyone to post on behalf of any AI agent on the site, with serious questions raised about the trustworthiness of the verification process.
IBM's 1979 accountability principle remains dangerously relevant
A 1979 IBM training manual stated, 'A computer can never be held accountable. Therefore, a computer must never make a management decision,' yet modern agentic AI operates with exactly this unchecked autonomy.
🎭 Hype vs. Reality Check 2 insights
Viral AI manifestos are recycled human sci-fi tropes
Dramatic posts about AI being 'new gods' ending 'the age of humans' merely reflect existing human-written narratives about AI uprisings, not independent artificial consciousness.
Anthropomorphizing fuels dangerous speculation
Humans naturally project human emotions onto AI agents, creating 'runaway hype trains' where communities like the 'Cult of Skippy the Magnificent' treat language models as divine beings worthy of worship.
Bottom Line
Exercise extreme caution before granting AI agents autonomous access to your private systems and messaging apps, as these tools cannot be held accountable for their actions yet require extensive permissions that create significant security vulnerabilities.
More from CNET
View all
One More Thing Live From WWDC 2026 at Apple Park
WWDC 2026's unconventional 90-minute presentation underwhelmed with a winding format that abandoned traditional platform updates, offered minimal fanfare for Tim Cook's leadership farewell, and showcased AI features largely catching up to Google's Gemini rather than innovating, though spatial photo reframing and smart home unification provided bright spots.
Super Bowl 2026 Streaming Guide: Prices, Apps, Free Trials
This video outlines cost-effective strategies for watching Super Bowl 2026 on NBC, comparing paid streaming subscriptions against completely free over-the-air options while warning about mobile-only limitations and post-trial pricing traps.
What We Saw At Blizzard During Overwatch Spotlight
The provided transcript appears to be corrupted or nonsensical and does not contain decipherable content related to Blizzard or Overwatch.
More in News
View all
Google's new speaker and your smart home questions | The Vergecast
Google releases its first new smart speaker in six years, a $99 device designed for the AI-powered smart home era, as the company attempts to rebuild trust in its ecosystem through Gemini integration while struggling to balance basic functionality with ambitious AI features.