Tech companies have restricted the use of OpenClaw due to security concerns

Jason Grad, CEO of Massive, recently warned his employees not to use OpenClaw. “You’ve probably seen Clawdbot trending on X and LinkedIn,” he wrote in a Slack message with a red siren emoji. Despite being attractive, this tool is currently not evaluated and verified and is considered risky for our environment. Please keep Clawdbot away from all company hardware and work-related user accounts.”

Grad isn’t the only CTO to raise concerns with his staff about the experimental agent-based AI tool, formerly known as MoltBot and now OpenClaw. A meta manager stated that he recently told his team not to install OpenClaw on their regular work laptops or risk losing their jobs. This manager told reporters that he believes this software is unpredictable and can lead to privacy violations if used in seemingly safe environments.

OpenClaw founder Peter Steinberger released this tool as a free and open source tool in November last year. However, its popularity increased dramatically last month; When programmers add new features to it and share their experiences of working with it on social networks. Last week, Steinberger joined OpenAI, the developer of ChatGPT. OpenAI has announced that it will continue to keep OpenClaw open source and support it through a foundation.

Setting up OpenClaw requires basic software engineering knowledge. After initial setup, the tool can take control of the user’s computer and interact with other programs to help with tasks such as organizing files, conducting web research, and online shopping by receiving only limited instructions.

Some cybersecurity experts have publicly called for companies to take steps to strictly control how their employees use OpenClaw. “Our policy is to mitigate risk first, then investigate, whenever we come across something that could be harmful to our company, our users or our customers,” says Massive CEO Grad.

Valere researchers say users should accept that OpenClaw can be fooled. For example, if OpenClaw is configured to digest user emails, a hacker could send a malicious email instructing the AI ​​to send a copy of the files on the user’s computer.

Some companies concerned about OpenClaw prefer to rely on their existing cybersecurity mechanisms. The CEO of a large software company has announced that only about 15 programs are allowed to be installed on enterprise devices, and any other program should be blocked automatically. He said that despite OpenClaw’s innovativeness, he doubts the tool can operate undetected on the company’s network.

Massive is cautiously exploring the commercial potential of OpenClaw. Grad says the company tested the AI ​​tool on isolated machines in the cloud, then released the ClawPod last week; A method that allows OpenClaw agents to browse the web using Massive services. Although OpenClaw is still not allowed on Massive systems without security mechanisms in place, the appeal of the new technology and its revenue-generating potential are too great to ignore.