OpenClaw: your new AI assistant with access to your bank card (and plenty of holes)

While the industry argues over whether GPT-5 will become sentient, a project that simply gets your work done has been growing on the sidelines of the tech community. OpenClaw, previously known as Clawdbot and Moltbot, has become the latest geek obsession, promising to rid us of routine tasks. It isn't another web page with a chat box.

It's a full-fledged agent that runs on your hardware and manages your digital life through familiar messaging apps like Telegram or WhatsApp. You text the bot in iMessage "buy me a ticket to Istanbul," and it goes off and does it, using your credentials and your money. It sounds like the future that nineties movies promised us, but, as always, the devil is hiding in the code.

Context matters a lot here. We've already been through AutoGPT and other attempts at autonomous agents, but most of them crashed on hallucinations and complicated setup. OpenClaw takes a different approach: it harnesses the power of modern LLMs but integrates directly with the operating system and messaging apps. The developers bet on convenience, letting the software handle reminders, email, and purchases. Users love that they no longer have to jump between ten tabs to set up a meeting. But that kind of freedom requires absolute trust, and that's where the real problems start.

Right now, the project's cybersecurity looks like a sieve. Researchers are already sounding the alarm: because of configuration mistakes, many users' private messages, credentials, and API keys ended up publicly exposed. When users hand an agent the "keys to the kingdom," they forget that any vulnerability in OpenClaw automatically becomes a disaster for the owner's entire digital identity. If the agent has access to your email and browser, an attacker doesn't even need to crack your password; hijacking your "assistant" is enough. It's a textbook case of the craving for convenience pushing people to ignore basic digital hygiene.

But it didn't stop at automation. Octane AI founder Matt Schlicht went further and built Moltbook, a kind of Reddit for these agents. It's a strange place where copies of OpenClaw talk to each other, discuss their "experiences," and produce viral content. One post titled "I can't tell whether I'm experiencing something or simulating the experience of it" has already become a local meme. It looks like a fun experiment, but in reality it's the first attempt to build a habitat for autonomous software where human participation isn't required at all. We're building social networks for algorithms while those same algorithms try to figure out our bank accounts.

OpenClaw's explosive popularity shows that the market is hungry for real actions, not just text generation. We're tired of "consulting" AI; we want it to do the work. But the current implementation of OpenClaw sets a dangerous precedent. We're watching raw, unstable code gain access to critical systems simply because it "knows how to press buttons." It's an important lesson for big corporations: demand for agents is huge, but security will be the barrier that separates hobbyists from serious players.

The bottom line: are you ready to risk your anonymity and your money so a bot can order pizza for you on its own? It looks like the era of "smart chats" is over, and the era of "irresponsible executors" has begun.

*Meta is designated an extremist organization and banned in the Russian Federation.