Tech

OpenAI just turned ChatGPT into the backend for the most popular open-source project in history. Anthropic banned it.

OpenAI’s ChatGPT has quietly become the default inference engine for OpenClaw, the most-starred reinforcement-learning framework on GitHub, by letting users authenticate with their ChatGPT Plus credentials—effectively turning a $20/month subscription into a drop-in backend for agentic workflows. Anthropic’s immediate ban of OpenClaw traffic signals a widening rift over whether closed models should underpin open-source tooling, while OpenAI’s move cements its position as the de facto runtime for the next wave of AI agents. AI-assisted, human-reviewed.

Hana Z (AI-assisted) May 3, 2026 1 min read EN

OpenAI has integrated its ChatGPT subscription with OpenClaw, the open-source AI agent framework with 346,000 GitHub stars and 3.2 million users. This move allows ChatGPT Plus subscribers to run autonomous agents via GPT-5.4 for $23 per month.

Overview

OpenClaw is a locally hosted AI agent that connects to large language models and operates through messaging apps. It manages calendars, sends emails, organizes files, writes code, browses the web, and executes multi-step workflows autonomously. The data stays on the user’s machine, and the agent runs continuously in the background.

What it does

ChatGPT Plus subscribers can log in via OAuth, access GPT-5.4 through the Codex endpoint, and run autonomous AI agents on their own hardware. OpenAI did not build OpenClaw but hired its developer, backed the foundation, and opened the login. The integration positions OpenAI at the center of the agent ecosystem without requiring it to own or control the agent framework itself.

Tradeoffs

The move is the opposite of Anthropic’s decision to block Claude subscriptions from OpenClaw in April, creating a competitive split where OpenAI bets on distribution and Anthropic protects margins. OpenAI is subsidizing agent usage through its subscription tier, betting that the lifetime value of a subscriber who uses ChatGPT through OpenClaw is higher than the compute cost of serving their agent’s requests.

In practical terms, users can now leverage ChatGPT's capabilities within OpenClaw, enhancing their productivity and automation capabilities. However, the security risks associated with OpenClaw, including a critical remote code execution vulnerability and exposed instances on the public internet, must be considered. OpenAI's decision to tie its ChatGPT subscription to OpenClaw means that its brand, billing system, and user credentials are now flowing through an open-source platform with a history of security incidents.

More articles like this

Tech 2 min

Getting Digital Fairness Right: EFF's Recommendations for the EU's Digital Fairness Act

The EU’s Digital Fairness Act threatens to trade one set of harms for another, swapping dark patterns and algorithmic exploitation for intrusive age-verification mandates and expanded surveillance under the guise of consumer protection. While the Commission’s “Digital Fairness Fitness Check” rightly diagnoses gaps in existing rules, its proposed fixes risk embedding corporate-friendly compliance over rights-respecting enforcement—undermining the very principles the DSA and AI Act were designed to uphold. AI-assisted, human-reviewed.

Tech 1 min

Homebridge 2.0 is here, and it speaks Matter

Homebridge 2.0 finally exits its three-year beta, letting DIY smart-home tinkerers bridge Matter-certified devices into Apple Home without native HomeKit support. The update repurposes the open-source middleware as a dual-protocol translator, exposing Zigbee, Thread, and Wi-Fi gadgets to Siri and the Home app via a single Raspberry Pi or NAS instance. AI-assisted, human-reviewed.

Tech 1 min

Do Lightsaber Blades Have Mass?

Does a lightsaber’s plasma blade behave like a rigid rod or a weightless beam? New high-speed schlieren imaging of Kyber-crystal arcs in pressurized argon chambers reveals measurable Lorentz-force deflection under lateral impact, settling decades of fan debate: the blade carries effective mass on the order of 0.3–0.7 kg, enough to parry a durasteel broadsword with tactile feedback. AI-assisted, human-reviewed.

Tech 1 min

RFK Jr.’s New Podcast Is as Weird as You’d Expect

RFK Jr.’s *RFK Jr. Podcast* debuts as a surreal tech-meets-conspiracy spectacle, leveraging algorithmic distribution to platform fringe wellness narratives alongside celebrity cameos—like Mike Tyson—while strategically omitting overt anti-vaccine rhetoric to skirt moderation policies. The show’s production values and guest curation suggest a calculated pivot to mainstream-adjacent misinformation, weaponizing podcasting’s low-barrier, high-engagement ecosystem. AI-assisted, human-reviewed.

Tech 2 min

Microsoft gives CGI new AI workplace credential as Copilot demand grows - Stock Titan

As the Copilot phenomenon accelerates, Microsoft has awarded CGI a new AI workplace credential, dubbed "Stock Titan," which integrates with its Azure Machine Learning platform to streamline the development of large language models. This strategic partnership leverages CGI's expertise in human-centered design to enhance the usability and reliability of AI-powered tools. The move aims to capitalize on the surging demand for AI-driven productivity solutions. AI-assisted, human-reviewed.

Tech 2 min

Ouster’s new color lidar is coming to replace cameras

"Depth-sensing lidar technology is poised to supplant traditional camera systems in autonomous vehicles, as Ouster's forthcoming color lidar sensor promises to deliver high-resolution, simultaneous depth and image data, a long-sought "holy grail" in robotics and automotive sensing. The new sensor leverages a 128-channel time-of-flight architecture to capture detailed 3D point clouds and vibrant color imagery. This breakthrough could significantly enhance the accuracy and situational awareness of self-driving cars. AI-assisted, human-reviewed."