Ten seconds. That’s how long it took crypto scammers to seize both the old GitHub organization and X account after Peter Steinberger released them during a forced rebrand. The Austrian developer, who founded PSPDFKit before selling to Insight Partners, had just built one of the fastest-growing open-source AI projects on the internet. Clawdbot had amassed over 60,000 GitHub stars as a self-hosted AI agent that could actually execute tasks, not just chat. On January 27, 2026, that momentum hit a wall.
“I was forced to rename the account by Anthropic,” Steinberger wrote on X. “Wasn’t my decision.”
The problem was the name. Clawdbot’s mascot, a space lobster named “Clawd,” phonetically matched Anthropic’s flagship AI product, Claude. When Steinberger asked if he could simply drop the “d” and rebrand to “Clawbot,” the answer came back: “Not allowed to.” Even near-phonetic similarity was unacceptable. The project is now called Moltbot, with a new mascot named Molty. “Molt fits perfectly,” Steinberger explained. “It’s what lobsters do to grow.”
Three days before the rebrand, Steinberger had appeared on a podcast confidently stating he believed the name was “legally viable.” He said he had checked and found “no trademark for this.” His search missed the conflict that would force him to abandon a brand 60,000 developers already knew.
Why “Clawd” Creates Trademark Problems for Anthropic
In American and Canadian English, “Claude” and “Clawd” are phonetically identical. Say them aloud. The sounds are indistinguishable. Trademark law treats this phonetic equivalence as a serious problem regardless of spelling differences.
Courts have consistently found marks with similar phonetic profiles create likelihood of confusion. The Trademark Trial and Appeal Board found SEYCOS confusingly similar to SEIKO for watches. CRESCO was too close to KRESSCO. ENTELEC infringed on INTELECT. These cases establish a clear principle: if consumers might hear two marks and confuse them, the second user has a problem. Spelling creativity does not save you. The legal standard recognizes that consumers encounter trademarks aurally as often as visually, through conversations, podcasts, video content, and word of mouth.
The likelihood of confusion test examines appearance, sound, meaning, and commercial impression. The critical misconception many developers have is thinking the test asks whether marks can be distinguished side by side. It does not. The test asks whether consumers encountering the marks separately, in different contexts, might confuse the source. Someone hearing about “Clawdbot” and “Claude” might reasonably wonder if Anthropic made both.
Anthropic filed its CLAUDE trademark on February 10, 2023. The company now holds 27 trademarks and has demonstrated willingness to enforce them. Trademark holders who fail to police their marks risk dilution claims and potential abandonment arguments. When a viral open-source project with 60,000 stars starts using a phonetically identical name, the trademark owner has to act.
This is where Steinberger’s self-conducted search failed him. He checked for “Clawdbot” as a trademark. The name wasn’t registered. But he missed the phonetic conflict with an established mark. Conducting a professional trademark clearance search catches these phonetic overlaps that database searches for exact matches miss.
Trademark Tensions in the AI Gold Rush
The Clawdbot situation reflects broader tensions between AI companies and the developer ecosystems building around their products. Just weeks earlier, on January 9, 2026, Anthropic had blocked third-party tools from spoofing Claude Code access. DHH, creator of Ruby on Rails, called Anthropic’s recent moves “customer hostile.”
The developer community split. Some argued Anthropic had every right to protect its trademark and API access. Others saw overreach, pointing out that Clawdbot used the official API legitimately and simply happened to have an unfortunately similar name. The project was not pretending to be affiliated with Anthropic. Steinberger was not selling a competing AI model. He built a tool that made Claude more useful, and Anthropic still forced him to change the name.
This tension will only intensify as AI tools proliferate. “Claude” derives from a common name, which creates a broad phonetic protection zone. Cloud, Clod, Claw, Clawd: variations that might sound innocent become trademark landmines when a major tech company has already staked its claim. Developers building AI tools face a naming landscape increasingly cluttered with protected terms.
For open-source developers specifically, building on proprietary AI platforms creates naming vulnerability. Your project’s visibility becomes a liability. At 60,000 stars, Clawdbot was impossible to ignore. An aggressive trademark protection strategy from a well-funded AI company will find you. Obscurity provides temporary protection. Success eliminates it.
Some developers are responding by moving toward alternatives. OpenAI’s Codex CLI, released under Apache 2.0 license, has attracted developers wary of building too deeply on any single company’s infrastructure. The Clawdbot rebrand became a cautionary tale shared in AI development communities within hours of Steinberger’s announcement. Developers on Hacker News and Reddit debated whether building on any proprietary AI platform was worth the risk when trademark enforcement could force a rebrand at any moment.
Protecting Your Software Brand Before Launch
Steinberger’s experience illustrates why professional trademark searches catch conflicts that self-searches miss. He searched for “Clawdbot” and found nothing. A trademark attorney would have searched phonetic variants of established marks in the AI space and flagged “Claude” immediately.
Phonetic analysis becomes essential when choosing names derived from common words or names. “Claude” sounds like claw, clod, cloud. The phonetic danger zone extends further than most developers realize. When I conduct trademark clearance for software projects, I search not just the proposed mark but every plausible pronunciation, common misspelling, and phonetic neighbor.
Tech companies are actively enforcing their trademark portfolios. Anthropic’s 27 trademarks represent substantial investment in brand protection. Google, Meta, Microsoft, and OpenAI have similarly built trademark arsenals around their AI products. Assuming you can slide under the radar because your project is open source is a miscalculation.
Open source does not exempt you from trademark law. The GPL and MIT licenses govern code distribution. Trademark law governs brand identity. You can fork someone’s code legally while still infringing their trademark. Steinberger’s 60,000 stars meant 60,000 potential witnesses to the alleged infringement. Success made him a target. Every blog post mentioning Clawdbot, every tutorial video, every GitHub discussion reinforced the connection between that name and AI agent functionality in the same space as Claude.
The rebrand logistics deserve attention too. Steinberger attempted to rename both his GitHub organization and X handle simultaneously. In the seconds between releasing the old names and claiming new ones, scammers struck. If you must rebrand, secure your new identity first, coordinate the transition carefully, and consider that bad actors monitor high-profile accounts for exactly these opportunities. The scammers who seized Clawdbot’s old accounts immediately began promoting cryptocurrency schemes to Steinberger’s confused followers.
The cost comparison makes the case clearly. A trademark clearance search might cost a few hundred dollars. Securing federal trademark registration early costs $225 to $600 per class. Steinberger’s forced rebrand at 60,000 users cost incalculable brand equity, community confusion, security compromises from scammer-seized accounts, and the time to rebuild recognition under a new name.
Act Before Your Brand Goes Viral
Steinberger’s brand exploded before he secured trademark protection. At 60,000 stars, a forced rebrand means losing everything you built under the original name. The irony is painful: his success made him a target. A smaller project might have continued unnoticed for years.
For AI developers and software startups, the lesson is clear. Establish trademark rights before launch. Professional phonetic analysis prevents the Clawd/Claude problem. Registration creates leverage for negotiation rather than forced compliance. If Steinberger had owned a trademark for Clawdbot, the conversation with Anthropic would have been different, potentially involving coexistence agreements rather than demands.
If you are building a software project and choosing a brand name, a trademark attorney can identify phonetic conflicts before launch and help you secure rights to your chosen name. Contact me for a consultation before your project gets big enough to attract enforcement attention.
