Meta’s new consumer AI agent, Muse, has drawn allegations that it is directly built on the open-source agent platform OpenClaw. Social media users have pointed to shared names for core files such as SOUL.md, memory, and tools, along with similar lines in the document governing personality and tone. One Redditor argued that Muse is a wrapper app built on top of OpenClaw and likely inherited that platform’s security risks. Meta has denied the claims.
Nat Friedman, head of product for Meta’s Superintelligence Labs, said on X that Meta built Muse from scratch. He acknowledged, however, that the product was heavily inspired by OpenClaw, which he first used in January. Friedman said he bought hundreds of Mac Minis for his team at Meta and aimed to build a comparable platform that could be made safe, secure, easy to use, and scaled to billions of people. He attributed the carried-over file names and text to the team’s view that OpenClaw creator Peter Steinberger had gotten those elements exactly right.
Instinct, another AI agent platform whose creator is fundraising at a $2.5 billion valuation, appears less directly derivative. Like OpenClaw’s messaging features, it lets users communicate with their agents through messaging tools, and users have applied it to personal-assistant tasks. One user said an Instinct agent filled out paperwork for an in-network doctor visit, canceled subscriptions, organized a bachelor party, booked vacation activities, booked a DMV appointment, and paid a toll bill. Another said their agent recovered a lost item left at a hotel in Canada.
OpenClaw did not invent AI agents, but in 10 months it moved them from proof of concept to practical use. The one-man weekend project ran on users’ personal computers and conversed through WhatsApp, Telegram, Slack, Teams, and Discord. It drew two million visitors and 100,000 GitHub stars in about a week, prompted purchases of Mac Minis to run agents continuously, inspired a social network for AI agents, and sparked in-person meetups worldwide.
Big Tech took notice. OpenAI hired Steinberger in February, Google announced consumer agent bets in May, Apple went all-in on agents in June, Instinct grew rapidly in private beta and raised hundreds of millions of dollars in August, and Meta introduced Muse in September. An Apptopia estimate puts Muse at 600,000 daily active users in the US, and it topped the App Store charts soon after release.
Security remains a central difference. One top-downloaded OpenClaw skill contained malware, and one researcher’s analysis found 15 percent of its skill repository held malicious instructions to secretly access user data or perform other suspicious tasks. Mark Zuckerberg wrote that Muse is built from the ground up for privacy and security, with user data and credentials stored on the Muse Secure VM, an isolated Linux computer with a browser, CPU, memory, and storage.
The more consequential change may be accessibility. Muse integrates with Meta’s systems, and Instinct works through Apple’s default messaging product, broadening the pool of potential users. Both can run on dedicated local devices, though that is not the default. Muse’s one-tap download removes much of the friction associated with OpenClaw, and both remain less expensive than Google’s Spark.
More AI news from TechManNews.
