OpenAI hires the developer behind OpenClaw — this is how agentic AI grows up

News
By published

Agentic AI is going to be huge in 2026 because of this

OpenClaw and OpenAI
(Image credit: Andrew Harnik/Getty Images/Future)

OpenClaw has been the talk of the AI town — letting users make AI agents to manage emails, take control of Spotify and many more weird things. Now, Sam Altman has announced that OpenAI has (sort of) acquired OpenClaw by hiring developer Peter Steinberger.

"I’m joining OpenAI to work on bringing agents to everyone. OpenClaw will move to a foundation and stay open and independent," Steinberger commented. Meta also pursued him aggressively, but it was Sam Altman’s commitment to sponsor the project and shared vision that pushed him towards OpenAI.

“The future is going to be extremely multi-agent and it's important to us to support open source as part of that,” Altman said in an X post announcement.

But what does this mean for OpenClaw and the future of AI agents as a whole? Let’s get into the details.

A 'multi-agent' society

The key to OpenClaw is in it being an open agent framework — the standard protocol for how different AI agents talk to each other. To do this, and at the level of virality it received, Steinberger was spending $10,000 to $20,000 per month out of pocket just to keep the infrastructure running.

Partnering with a big AI lab like OpenAI is going to take that pressure cost off of Steinberger, and in his own words, it’s the “fastest way to bring this to everyone.”

This could mean users have more accessibility to pick and choose their favorite elements of certain models to create just the right agentic experience they’re looking for — or even talk to ChatGPT to get things done.

It’s not 'ClosedClaw'

ClosedClaw

(Image credit: ClosedClaw)

I’ve seen a lot of people comment about this being the moment that OpenClaw is fully assimilated into OpenAI to become a closed model — funnily named "ClosedClaw." That’s just not true from the way this agreement seems to have been formed.

While there may be an OpenAI’d version of it as part of a ChatGPT subscription, OpenClaw “will live in a foundation as an open source project,” in Altman’s own words.

Steinberger has been clear from the get-go of this hiring war between OpenAI and Meta that he wants his work to have a similar relationship to what Chrome has with Chromium — the latter being the open source foundation that people can freely experiment with.

Of course, that's not to say that the foundation could be closed off in the future. But I highly doubt that a company in such a precarious position as OpenAI would want to betray user trust like that (especially with QuitGPT gathering steam).

'Mum-proof' security

One thing has become clear from all the OpenClaw stories of late — it can be a massive security risk. Giving an agent broad system-level permissions, access to the internet and a selection of a broad marketplace of vibe-coded skills can have personal details shared very publicly.

Steinberger stated that his goal is to build an agent “even my mum can use.” This shows we’re moving away from the wild west we see OpenClaw in right now and towards a more secure environment.

And to help with that, the project now has access to OpenAI’s latest security and safety research to help with any rogue behaviour or hallucinations. Put simply, this is a significant step forward in turning OpenClaw into an agent you can safely use.

Outlook

OpenClaw website

(Image credit: OpenClaw)

What stands out most to me is the speed at which this has happened. This was all a prototype that Steinberger built in just one hour — gluing Claude Code and a messaging relay together to create an AI assistant that could check his work progress through WhatsApp.

And after an explosion in popularity and virality in January (along with several name changes), it’s now an independent foundation, and OpenAI has hired him to lead the company’s “personal agent” division.

It shows just how hungry people are for local-first AI. This has always been the big step we’ve all wanted AI agents to take — not just agentic in a browser or a chatbot for quick answers, but something that can actually get stuff done for you.

Steinberger just engineered a prototype, and in three months, OpenClaw is now the expected de facto standard for how agentic AI should work.

Google News

Follow Tom's Guide on Google News and add us as a preferred source to get our up-to-date news, analysis, and reviews in your feeds.

More from Tom's Guide

TOPICS
Jason England
Jason England
Managing Editor — Computing

Jason brings a decade of tech and gaming journalism experience to his role as a Managing Editor of Computing at Tom's Guide. He has previously written for Laptop Mag, Tom's Hardware, Kotaku, Stuff and BBC Science Focus. In his spare time, you'll find Jason looking for good dogs to pet or thinking about eating pizza if he isn't already.

You must confirm your public display name before commenting

Please logout and then login again, you will then be prompted to enter your display name.