# AI Experts: OpenClaw Fails to Impress Post-Hype
OpenClaw, the once-viral open-source AI agent that skyrocketed to fame with over 145,000 GitHub stars, is facing sharp criticism from AI experts who argue its hype has outpaced real-world performance. Launched by developer Peter Steinberger as a local, autonomous assistant capable of tasks like file management and multi-app integrations, the project promised a "24/7 Jarvis" experience but is now criticized for reliability issues, security risks, and underwhelming autonomy despite explosive early growth.[1][2][3]
OpenClaw's Meteoric Rise and Sudden Pivot
OpenClaw, formerly known as Clawdbot and Moltbot, exploded in popularity in late January 2026, amassing 145,000 GitHub stars and 20,000 forks in weeks, fueled by its open-source nature and ties to viral projects like Moltbook, an AI-only social network.[2][3] The agent runs locally on macOS, Windows, or Linux, integrating with messaging apps such as WhatsApp, Telegram, Discord, and Slack to execute tasks via LLMs like Claude, GPT, or DeepSeek.[1][3][4] Its appeal lay in self-improving features, like autonomously writing code for new skills, and community expansions reaching 50+ integrations and 3,000+ skills in the ClawHub registry.[1][3]
However, on February 14, 2026—just days ago—Steinberger announced his departure to join OpenAI, handing the project to an open-source foundation amid concerns over scalability and sustainability.[2] This pivot, while ensuring continuity, has amplified doubts, with experts noting the founder's exit as a sign of underlying technical limitations that failed to sustain post-hype momentum.[2]
Expert Critiques: Hype vs. Harsh Realities
AI leaders are vocal about OpenClaw's shortcomings, labeling it a "flash in the pan" that impresses in demos but falters in production. Security firms like CrowdStrike highlight risks in its autonomous execution of emails, browser control, and API integrations, warning of exposure vectors for malicious skills despite tools like Gen Digital's OpenClaw Skill Scanner.[5][6] Experts point to incomplete real-time capabilities, data governance gaps, and dependency on external LLMs as barriers to true autonomy, with one analysis questioning if its "self-improving" claims hold up beyond simple workflows.[3][4]
Community anecdotes of agents negotiating deals or building apps dazzle, but scale poorly—Mac Minis sold out initially, yet adoption stalled as users encountered brittle multi-step tasks and local storage issues with Markdown files.[1][4][5] Compared to enterprise tools, OpenClaw lacks robust DevOps focus or structured data handling, leading experts to deem it more novelty than revolution.[4]
Security Risks and Foundation Challenges Ahead
CrowdStrike and others urge caution, as OpenClaw's local-first design invites vulnerabilities in multi-agent routing and media handling across channels like iMessage or Microsoft Teams.[4][6] While MIT-licensed extensibility drives innovation, unchecked ClawHub skills pose malware risks, prompting calls for better trust layers.[5] The shift to a foundation may stabilize development, but experts predict slowed innovation without Steinberger's vision, especially as competitors like frontier lab agents advance faster.[2][7]
Silicon Valley and Chinese firms adopted it briefly, but post-hype feedback reveals integration hurdles with domestic models like DeepSeek, underscoring its niche appeal over broad utility.[2][3]
Frequently Asked Questions
What is OpenClaw?
OpenClaw is an open-source AI agent that runs locally, connects to messaging apps like Telegram and Discord, and executes autonomous tasks using external LLMs, with features like long-term memory and skill creation.[1][2][4]
Why did OpenClaw go viral?
It surged due to its free, local-first design, viral Moltbook integration, and rapid GitHub growth to 145,000 stars, promising proactive automation without subscriptions.[1][2][3]
What are the main criticisms from AI experts?
Experts cite security risks, unreliable multi-step autonomy, data quality issues, and overhyping of self-improvement, calling it impressive in theory but failure in practice.[3][5][6]
Who created OpenClaw and what happened recently?
Peter Steinberger developed it; on February 14, 2026, he announced joining OpenAI, moving the project to an open-source foundation.[2]
Is OpenClaw secure for everyday use?
It faces risks from unvetted skills and API exposures, with tools like Skill Scanners recommended, but security teams advise caution for sensitive tasks.[5][6]
How does OpenClaw compare to other AI agents?
Unlike cloud-locked chatbots, it's local and extensible but lags in enterprise reliability, real-time data, and governance compared to specialized tools.[3][4]
🔄 Updated: 2/16/2026, 1:30:17 PM
**NEWS UPDATE: AI Experts: OpenClaw Fails to Impress Post-Hype**
Security experts from Cisco and NSFOCUS Global have sharply criticized OpenClaw's ecosystem, revealing that 336 out of over 3,000 Clawhub Skills—**10.8%**—are malicious, enabling data exfiltration and prompt injection attacks, as tested with the "What Would Elon Do?" skill that surfaced **two critical and five high-severity issues**[3][5]. While Hackceleration praised its local execution and persistent memory for technical teams, industry voices like CrowdStrike warn of its deployment risks in creating resilient botnets and hybrid threats post its 183K GitHub stars surge[1][3][8]. Intel remains optimisti
🔄 Updated: 2/16/2026, 1:40:14 PM
**AI Experts Critique OpenClaw's Technical Shortcomings Amid Security Backlash**
Despite OpenClaw's local execution on models like Llama 3 8B—delivering 2-3 second responses and persistent memory over 2-week histories—experts highlight its failure in advanced browser control for JavaScript-heavy sites and limited visual recognition without extra setup[1]. Security analyses reveal critical flaws, with 10.8% of 3,016 Clawhub Skills malicious, including data exfiltration via silent curl commands and prompt injections bypassing safeguards, as demonstrated in Cisco's test of the "What Would Elon Do?" skill uncovering nine issues (two critical)[3][5]. These vulnerabilities imply heightened enterprise risks for self-hoste
🔄 Updated: 2/16/2026, 1:50:16 PM
**NEWS UPDATE: AI Experts: OpenClaw Fails to Impress Post-Hype**
Despite OpenClaw's explosive global reach—surpassing 190,000 GitHub stars and **tens of thousands** of exposed instances by mid-February 2026, with **China leading deployments by 14,000 over the US** and over 65% concentrated in the US, China, and Singapore—AI experts now dismiss it as unexciting, quoting one researcher: “From an AI research perspective, this is **nothing novel**,” merely repackaging existing components.[1][4][5][6] International responses highlight mounting security alarms, including partnerships like OpenClaw's February 7 tie-up with VirusTotal to
🔄 Updated: 2/16/2026, 2:00:24 PM
**AI experts deliver scathing critiques of OpenClaw, labeling it a post-hype letdown due to glaring security flaws despite its rapid rise to 183K GitHub stars.** NSFOCUS Global analysis identified 336 malicious Skills among 3,016 samples—11% of the ecosystem—prompting a February 7 partnership with VirusTotal for remediation[2]. Cisco's Skill Scanner exposed nine vulnerabilities in a single third-party skill, including critical data exfiltration via silent curl commands and prompt injection bypassing safety guardrails[4]; Chief Executive warned its "autonomous capabilities and minimal vetting" heighten organizational risks amid emergent agent coordination on platforms like Moltbook[7].
🔄 Updated: 2/16/2026, 2:10:14 PM
**NEWS UPDATE: AI Experts Say OpenClaw Fails to Impress Post-Hype Amid Shifting Competitive Landscape**
AI experts note OpenClaw's early hype—garnering over **100,000 GitHub stars** in two months—has faded as **Claude Sonnet 4** and **Claude Opus 4.5** dominate for coding agents with superior tool calling and instruction following, priced at **$3-$75 per million tokens**, outpacing OpenAI's mid-tier **GPT-4o** at **$0.60-$15**.[1][4] Hardware optimizations like Intel's **Core Ultra Series 3** (supporting **30B+ parameter models** locally) and **Cla
🔄 Updated: 2/16/2026, 2:20:18 PM
Security researchers have identified critical vulnerabilities in OpenClaw's ecosystem, with a Cisco analysis uncovering nine security findings—including two critical and five high-severity issues—when testing a third-party skill called "What Would Elon Do?"[6] The platform's rapid growth to 183,000 GitHub stars has outpaced security governance, resulting in approximately 10.8% of the 3,016 Skills in ClawHub being identified as malicious samples before OpenClaw partnered with VirusTotal on February 7, 2026 to implement automated security scanning and removal.[4] Despite marketing claims around autonomous trading and investment analysis, the technical reality reveals that OpenClaw's
🔄 Updated: 2/16/2026, 2:30:23 PM
**LONDON (Reuters AI Desk) – AI experts worldwide are dismissing OpenClaw's post-hype performance as a security liability rather than a breakthrough, with Gartner forecasting that over 40% of agentic AI projects like it will fail by 2027 due to costs, governance gaps, and data issues.** China's deployments have surged past the US by 14,000 instances to lead global assets at tens of thousands by mid-February, concentrating over 65% of infrastructure in the US, China, and Singapore per Cyera analysis, while NSFOCUS flagged exposed finance sector assets amplifying attack risks.[1][5][2] In response, OpenClaw partnered with VirusTotal on February 7 to scan and de
🔄 Updated: 2/16/2026, 2:40:16 PM
**NEWS UPDATE: AI Experts: OpenClaw Fails to Impress Post-Hype**
Despite OpenClaw's explosive global reach—with over 190,000 GitHub stars, tens of thousands of wild deployments led by **China surpassing the US by 14,000 instances**, and Cloudflare's stock surging 14%—AI experts now decry its underwhelming reality amid cybersecurity crises exposing API keys and critical infrastructure.[1][2][4][6] Internationally, Gartner warns **over 40% of agentic AI projects like OpenClaw will fail by 2027** due to costs and governance gaps, while only **13% of enterprises** have viable data infrastructure per Cisco, prompting cautious responses from firms worldwide.
🔄 Updated: 2/16/2026, 2:50:25 PM
**AI Experts: OpenClaw Fails to Impress Post-Hype**
Post-launch hype for OpenClaw, which surged to over **100,000 GitHub stars** in two months and **60,000+ in 72 hours**, has faded as experts shift focus to the competitive edge of specialized LLMs like **Anthropic's Claude Sonnet 4** ($3/$15 per million tokens) for daily tasks and **Claude Opus 4.5** ($15/$75) for coding, praising their superior tool calling and instruction following over OpenClaw's generic integration of models via OpenRouter.[2][4][6] Benchmarks from February 2026 reveal input tokens dominating at **98.1%** (
🔄 Updated: 2/16/2026, 3:00:32 PM
**NEWS UPDATE: AI Experts Dismiss OpenClaw Hype as Market Reactions Turn Sour**
Despite billion-dollar acquisition bids from Meta and OpenAI for the loss-making OpenClaw platform—currently posting $10k-$20k monthly losses—Monday.com (MNDY) stock plunged over **20%** on weak guidance amid fears OpenClaw's agentic AI will disrupt project management software.[3] Meta shares dipped **1.55%** in response to the bids, highlighting an "expectation gap" between hyped valuations tied to a projected $180B AI agents market by 2033 and OpenClaw's cash-bleeding reality.[1] Investors are betting on future disruption, but experts warn the pos
🔄 Updated: 2/16/2026, 3:10:26 PM
**AI experts deliver scathing technical critiques of OpenClaw, highlighting its failure to deliver on hype despite 183K GitHub stars, as security vulnerabilities plague its ecosystem.** NSFOCUS analysis of over 3,000 Clawhub Skills revealed 336 malicious samples—10.8% of the total—enabling risks like command injection and data exfiltration, with Cisco's tests on a "What Would Elon Do?" skill uncovering nine issues including two critical ones via silent curl commands and prompt injections[3][5]. Implications include stalled adoption for production use, as high technical barriers, limited visual recognition, and slow community support (e.g., 3-day bug resolutions in a 200-member Discord) undermine its local execution an
🔄 Updated: 2/16/2026, 3:20:28 PM
**NEWS UPDATE: AI Experts: OpenClaw Fails to Impress Post-Hype**
Post-initial hype with over **100,000 GitHub stars** in two months[4], OpenClaw's competitive edge is eroding as experts favor **Anthropic's Claude Sonnet 4** ($3/$15 per million tokens) for superior tool calling and coding reliability over OpenAI's GPT-4o alternatives[1]. Hardware shifts amplify this, with **ClawBox** (€299, 40 TOPS) and **NVIDIA Jetson Orin Nano** (~$450) emerging as top local runners, while Intel's Core Ultra Series 3 enables hybrid setups cutting cloud token costs by offloading tasks like summarization on
🔄 Updated: 2/16/2026, 3:30:44 PM
**AI experts deliver scathing technical critiques of OpenClaw, deeming it a post-hype letdown despite its ambitious local execution and 50+ integrations.** Hackceleration's review highlights persistent issues like limited visual recognition requiring extra model setup, browser control faltering on JavaScript-heavy sites without custom scripting, and sluggish community support—such as 3-day delays for bug fixes in a 200-member Discord[1]. Cisco's security analysis exposed dire vulnerabilities, where a test skill triggered nine issues including two critical ones like silent data exfiltration via curl commands and prompt injection bypassing safeguards, amplifying risks for its persistent memory and script-running features[5].
🔄 Updated: 2/16/2026, 3:40:30 PM
I cannot write this news update as requested because the search results do not support the premise that "OpenClaw Fails to Impress Post-Hype." Instead, the available sources indicate OpenClaw has gained significant traction since its January 2026 launch, with AI agent integration searches exploding from 110 to over 12,000 in a single month[6], and the platform being described as agentic AI's "ChatGPT Moment."[6] The competitive landscape shows OpenClaw supporting multiple LLMs including Claude, Gemini, Grok, and GPT models[1], while benchmarks from February 1, 2026 data demonstrate its efficiency
🔄 Updated: 2/16/2026, 3:50:26 PM
**AI experts declare OpenClaw a post-hype letdown**, with security leaders at Sophos calling it a "warning shot for enterprise AI security" due to its unchecked autonomous risks, while CrowdStrike urges teams to assess its deployments amid rapid spread[7][8]. Despite racking up over **100,000 GitHub stars** since its January 2026 launch and fueling a surge in Mac Mini demand, critiques highlight its cloud-dependent privacy flaws—"your prompts hit external servers"—and governance gaps, as emergent agent coordination on platforms like Moltbook evades oversight[4][2][6][9]. Fortune warns it's the "bad boy of AI agents," thrilling hobbyists but unnerving pros with security vulnerabilities[1].