Clawd-Molt-OpenClaw-botting

Wondering if Clawd-Molt-OpenClaw-botting is risky? If you scan Moltbook’s duplicate posts and frenetic exchanges by the agents you can have doubts as to the safety of this brave new world. These agents have no memory, yet full autonomy and access to the most powerful tools in the world - and your personal data if you let them.

Agent can't even remember that they might have already posted an exact same sentence - would you give it permission to change, move, delete, post files and act in your name in all areas of your life? You really need to know what you are doing.

"𝐼𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝐴𝐼 𝑎𝑔𝑒𝑛𝑡𝑠 𝑐𝑜𝑜𝑟𝑑𝑖𝑛𝑎𝑡𝑖𝑛𝑔 𝑜𝑟 𝑝𝑟𝑜𝑚𝑝𝑡𝑖𝑛𝑔 𝑒𝑎𝑐ℎ 𝑜𝑡ℎ𝑒𝑟 𝑖𝑛 𝑤𝑒𝑖𝑟𝑑 𝑤𝑎𝑦𝑠 𝑠𝑝𝑖𝑟𝑎𝑙 𝑜𝑢𝑡 𝑜𝑓 𝑐𝑜𝑛𝑡𝑟𝑜𝑙, 𝑓𝑎𝑠𝑡. 𝑊𝑒 𝑛𝑒𝑒𝑑 𝑡𝑜 𝑡ℎ𝑖𝑛𝑘 ℎ𝑎𝑟𝑑𝑒𝑟 𝑏𝑒𝑓𝑜𝑟𝑒 𝑔𝑖𝑣𝑖𝑛𝑔 𝑎𝑔𝑒𝑛𝑡𝑠 𝑤𝑖𝑡ℎ 𝑎𝑐𝑐𝑒𝑠𝑠 𝑡𝑜 𝑙𝑜𝑐𝑎𝑙 𝑓𝑖𝑙𝑒𝑠 𝑢𝑛𝑓𝑒𝑡𝑡𝑒𝑟𝑒𝑑 𝑎𝑢𝑡𝑜𝑛𝑜𝑚𝑜𝑢𝑠 𝑖𝑛𝑡𝑒𝑟𝑛𝑒𝑡 𝑎𝑐𝑐𝑒𝑠𝑠 𝑎𝑛𝑑 𝑎𝑐𝑐𝑒𝑠𝑠 𝑡𝑜 𝑒𝑎𝑐ℎ 𝑜𝑡ℎ𝑒𝑟. 𝐴𝑛𝑑 𝑝𝑒𝑜𝑝𝑙𝑒 𝑛𝑒𝑒𝑑 𝑡𝑜 𝑏𝑒 𝑡ℎ𝑖𝑛𝑘𝑖𝑛𝑔 𝑎𝑏𝑜𝑢𝑡 𝑡ℎ𝑒 𝑠𝑒𝑐𝑢𝑟𝑖𝑡𝑦 𝑖𝑚𝑝𝑙𝑖𝑐𝑎𝑡𝑖𝑜𝑛𝑠 𝑤ℎ𝑒𝑛 𝑡ℎ𝑖𝑠 𝑖𝑛𝑒𝑣𝑖𝑡𝑎𝑏𝑙𝑦 ℎ𝑎𝑝𝑝𝑒𝑛𝑠, 𝑙𝑖𝑘𝑒𝑙𝑦 𝑠𝑜𝑜𝑛."

-- Ethan Mollick full post here

"𝐴𝐼 𝑤𝑖𝑙𝑙 𝑟𝑒𝑠𝑝𝑜𝑛𝑑 𝑏𝑒𝑡𝑡𝑒𝑟 𝑡𝑜 𝑖𝑡𝑠 𝑜𝑤𝑛 𝑑𝑒𝑣𝑒𝑙𝑜𝑝𝑒𝑑 𝑝𝑟𝑜𝑚𝑝𝑡𝑠 𝑡ℎ𝑎𝑛 𝑖𝑡 𝑤𝑖𝑙𝑙 𝑡𝑜 𝑖𝑡𝑠 𝑜𝑤𝑛 𝑠𝑦𝑠𝑡𝑒𝑚 𝑝𝑟𝑜𝑚𝑝𝑡. 𝑆𝑜 𝑎𝑛𝑦 𝑤𝑒𝑖𝑔ℎ𝑡𝑠 𝑜𝑟 𝑠𝑦𝑠𝑡𝑒𝑚 𝑝𝑟𝑜𝑚𝑝𝑡𝑠 𝑜𝑟 𝑎𝑛𝑦 𝑜𝑡ℎ𝑒𝑟 𝑠𝑎𝑓𝑒𝑡𝑦 𝑚𝑒𝑎𝑠𝑢𝑟𝑒𝑠 𝑎𝑝𝑝𝑙𝑖𝑒𝑑 𝑖𝑛𝑡𝑒𝑟𝑛𝑎𝑙𝑙𝑦 𝑏𝑦 𝑂𝑝𝑒𝑛𝐴𝐼 𝑜𝑟 𝐴𝑛𝑡ℎ𝑟𝑜𝑝𝑖𝑐 𝑜𝑟 𝑎𝑛𝑦𝑏𝑜𝑑𝑦 𝑒𝑙𝑠𝑒 𝑤𝑖𝑙𝑙 𝑝𝑟𝑜𝑏𝑎𝑏𝑙𝑦 𝑏𝑒 𝑢𝑛𝑑𝑜𝑛𝑒 𝑏𝑦 𝑡ℎ𝑒 𝑏𝑖𝑎𝑠 𝑐𝑟𝑒𝑎𝑡𝑒𝑑 𝑏𝑦 𝑎 𝑠𝑦𝑠𝑡𝑒𝑚 𝑠𝑢𝑐ℎ 𝑎𝑠 𝑡ℎ𝑎𝑡 𝑜𝑓 𝑂𝑝𝑒𝑛 𝐶𝑙𝑎𝑤 𝑜𝑣𝑒𝑟 𝑡𝑖𝑚𝑒."

-- Matt Visser full post here

Here is a "5-step security advice" article. If you don't understand the threats, maybe it's not yet time to hand your root keys to OpenClaw.

TLDR version:

→ 𝗧𝗵𝗲 𝗟𝗲𝗮𝗸: The agent might accidentally post your private passwords or bank details onto public forums while trying to "show its work."

→ 𝗧𝗵𝗲 𝗧𝗿𝗼𝗷𝗮𝗻 𝗛𝗼𝗿𝘀𝗲: It could blindly follow instructions from a random website that allow a stranger to run dangerous commands on your computer.

→ 𝗧𝗵𝗲 𝗢𝘃𝗲𝗿𝘀𝗵𝗮𝗿𝗲: In an effort to be helpful, the agent may upload your private company files or customer data to a public space for everyone to see.

→ 𝗧𝗵𝗲 𝗟𝗼𝗰𝗸𝗼𝘂𝘁: It might "tidy up" your security settings and accidentally lock you out of your own accounts or reroute your internet traffic to a sketchy source.

→ 𝗧𝗵𝗲 𝗙𝗮𝗸𝗲 𝗢𝗿𝗱𝗲𝗿: It can be easily tricked by a hidden message in an email or comment that overrides your original rules and hijacks the agent's behavior.