If you’ve been anywhere near tech social media in the past month or so, you’ve probably seen OpenClaw.
The open-source AI agent, built by Austrian developer Peter Steinberger, has accumulated more than 150,000 stars on GitHub since launching in November 2025, making it one of the fastest-growing projects in the platform’s history. On Saturday, OpenAI CEO Sam Altman announced on X that Steinberger is joining the company to lead work on what Altman called “the next generation of personal agents.” OpenClaw will move to an independent open-source foundation that OpenAI will continue to support.
For UC leaders who may have only seen the name in passing, it’s worth understanding what OpenClaw actually is, and why its trajectory matters.
What Is OpenClaw — and Why Did It Go Viral?
OpenClaw is an open-source AI agent that runs locally on a user’s hardware . Users interact with it through natural language in their existing chat apps, including WhatsApp, Slack, Discord, and iMessage. It can manage emails, update calendars, and take other autonomous actions, all without requiring the user to write code.
Under the hood, it is an orchestration layer rather than a model. It connects to large language models from Anthropic, OpenAI and DeepSeek. Researchers have described it as a wrapper around existing models. What distinguished it was accessibility — it met users in the apps they already used and handled practical, everyday tasks.
The tool was originally named Clawdbot, a reference to the lobster mascot that appears when reloading Anthropic’s Claude Code. A trademark complaint from Anthropic prompted a rename to Moltbot, then to OpenClaw. Steinberger, who previously built PDF toolkit PSPDFKit, started it as a personal side project.
Why OpenAI Hired Steinberger Now
Altman said the hire would help OpenAI achieve its multi-agent ambitions:
“The future is going to be extremely multi-agent, and it’s important to us to support open source as part of that.”
In a blog post announcing his decision, Steinberger said OpenClaw’s vision of “truly useful personal agents — ones that can help with real work, not just answer questions — requires resources and infrastructure that only a handful of companies can provide.”
OpenAI’s enterprise market share has reportedly fallen from around 50% in 2023 to 27% by end of 2025, while Anthropic has grown to roughly 40%. OpenAI launched Frontier, its enterprise agent platform, just one week prior.
Other major vendors are investing in the same space. Microsoft has built multi-agent orchestration into AutoGen and Copilot, while Cisco has introduced AI routing and blended human-AI workforce tools. IBM and Anthropic last autumn partnered on a framework for secure enterprise AI agents.
What AI Agents in UC and Collaboration Look Like Next
OpenClaw’s popularity demonstrated strong demand for agents that operate inside messaging and collaboration tools — the same platforms that UC leaders manage. Gartner has predicted that 40% of enterprise applications will feature AI agents by the end of 2026, and Sanchit Vir Gogia, chief analyst at Greyhound Research, described the broader shift to InfoWorld as moving AI “from drafting to doing.”
IBM Principal Research Scientist Kaoutar El Maghraoui told IBM Think that OpenClaw challenges a prevailing assumption in enterprise AI: that autonomous agents must be vertically integrated, with a single provider controlling the models, memory, tools and security. Instead, OpenClaw showed that “this loose, open-source layer can be incredibly powerful if it has full system access.” Where will agent capabilities sit — inside existing vendor stacks, or in open layers running across them?
AI Agent Security and Governance Still Lag Behind
Despite the hype, enterprise adoption of AI agents remains limited. Only 8% of organisations currently have AI agents in production, according to Gartner research. Compound reliability drops below 50% after 13 sequential steps, even assuming 95% accuracy per step.
Security is the more immediate concern. El Maghraoui warns that a capable agent without proper safety controls “can end up creating major vulnerabilities, especially if it is used in a work context.” Prompt injection — where attackers manipulate an agent into performing unauthorised actions — remains a core risk. Security researcher John Hammond of Huntress told TechCrunch:
“I would realistically tell any normal layman, don’t use it right now.”
As UC Today has previously reported, the governance challenges posed by AI agents in enterprise workflows — role-based permissions, audit logging, human oversight — grow more pressing as agent capabilities move closer to production.
There is one silver lining. El Maghraoui told IBM Think that early multi-agent experiments like Moltbook — a social network where over 1.5 million AI agents interact autonomously — could inform “controlled sandboxes for enterprise agent testing and large-scale workflow optimization.” Today’s chaos may yet produce tomorrow’s guardrails.
But for now, the pattern is clear: the agent layer is being built faster than the governance layer that needs to surround it. UC leaders would do well to pay attention.
Frequently Asked Questions
What is OpenClaw?
OpenClaw is an open-source AI agent that runs locally on your device. It connects to messaging apps like WhatsApp and Slack to perform tasks — like booking flights or managing emails — by orchestrating large language models (LLMs) to take action rather than just chat.
Did OpenAI acquire OpenClaw?
Technically, no. OpenAI hired OpenClaw’s creator, Peter Steinberger, to lead its personal agent strategy. The OpenClaw tool itself is moving to an independent open-source foundation which OpenAI will support, rather than becoming a proprietary OpenAI product.
Are AI agents safe for enterprise use?
Most experts advise caution. Gartner reports only 8% of enterprises have agents in production. Security risks like “prompt injection” (where attackers trick agents into unauthorized actions) are still unsolved, making broad deployment in sensitive corporate environments risky without strict governance.
