Key points
- OpenAI CEO Sam Altman confirmed the company will build a humanoid robot designed to navigate human-centric environments.
- Altman stated that developing the robot's cognitive "brain" is significantly more difficult than constructing its physical hardware.
- A demonstration of the technology is expected in 2027, with widespread deployment likely years later.
OpenAI intends to build a humanoid robot, but its leadership emphasizes that creating the intelligent system to control the machine is far more difficult than constructing the physical body.
How it unfolded
- OpenAI CEO Sam Altman discussed the project on the Sources podcast, explaining that the human form is logical because the world is designed for people.
- Altman confirmed the company will build the robot, stating that matching the human form factor is beneficial for interacting with doors, keyboards, and tools.
- In a separate interview with Fortune, Altman predicted a cool demo in 2027 but noted that robots walking the streets would take several more years.
Who is affected
According to Engadget, OpenAI is currently recruiting specialists to address the complex challenges of physical artificial intelligence. The company is hiring for roles in motion control, state estimation, and actuator control. These positions focus on developing the machine-learning infrastructure required to train systems that interact with the physical world. The recruitment effort signals a shift from purely digital models to embodied AI agents.
What happens next
OpenAI has not disclosed specific details about what the 2027 demo will showcase or what the humanoid will look like. The company has also not stated when the public might interact with the device. Altman indicated that the immediate focus is on solving the cognitive challenges rather than refining the hardware aesthetics. The timeline suggests that functional prototypes will remain internal or limited for the foreseeable future.
The background in plain language
Teams developing physical AI must solve perception, reasoning, and actuation simultaneously. Systems need to interpret sensor data, plan actions, and execute precise movements while correcting for real-world friction and unpredictability. This requires integrating visual recognition with motor control loops. Engineers typically simulate these environments first to train models safely before deploying them on hardware. Understanding this complexity helps explain why hardware progress often outpaces reliable autonomous behavior.
What to do and how to stay safe: OpenAI
- Monitor vendor announcements for updates on physical AI safety standards and certification processes.
- Evaluate current infrastructure for potential integration points with future robotic automation systems.
- Review data privacy policies regarding video and sensor data collection in physical environments.
- Stay informed about regulatory developments concerning autonomous agents in public spaces.
Step-by-step guide: Injection attacks examples: how prompt leaks break AI systems
General security guidance from the Circuit Observer newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
When will OpenAI release a humanoid robot?
OpenAI CEO Sam Altman expects a demo in 2027, but stated that widespread deployment will take several more years.
Why is OpenAI building a humanoid robot?
The company believes the human form is logical because the world, including doors and tools, is designed for human bodies.
What is the main challenge in building this robot?
According to Altman, figuring out the "brain" that makes the robot work is a much bigger challenge than building the physical body.
