Transparency, Regulations and Compliance
Project Goals
The goal of this project is to provide a pioneering example of the domestic coexistence between humans and humanoids. The aim is to inform the necessary policy which will be drafted, refined and enforced in the coming years. By completing this experience in an open, thoughtful and public fashion, the experience can inspire debate which will influence regulatory standards to encourage a smooth transition into the embrace of domestic embodied intelligence.
When bringing a humanoid robot to live in Vienna, there are a number of regulatory bodies and standards that could come into play when attempting to maintain both ethical transparency and legal compliance. These range from global certifications all the way down to local city initiatives. Prior to the arrival of the robot, I have endeavored to contact all of the relevant authorities to ensure that my research experiment is compliant with any existing regulations as well as to establish transparency and communication with the regulatory bodies currently operating in the field today.
Global
On the global level, there is the ISO or The International Organization for Standardization. Headquartered nearby in Switzerland, the ISO is an independent, non-governmental, international standard development organization composed of representatives from the national standards organizations of member countries. There are a number of different sectors covered by the ISO, one of which is IT and Related Technologies. It is under this sector that we find the ISO 13482:2014 Robots and robotic devices. This paper specifies requirements and guidelines for the inherently safe design, protective measures, and information for use of personal care robots, in particular the following three types of personal care robots:
mobile servant robot;
physical assistant robot;
person carrier robot.
As the end user, this reads more as an applied cautionary document rather than a certification process outline. The main interest for me within this document was the idea of the many types of classifications and exceptions that exist within the personal care robot designation. The exceptions to the classifications include robots travelling faster than 20 km/h, robot toys, water-borne robots and flying robots, industrial robots, which are covered in ISO 10218, robots as medical devices, military or public force application robots. Although Tova will be a mobile servant robot, I spent days stretching my imagination and wondering what would have to change with its physical capabilities and internal mechanisms to fall under an alternate designation.
European Union
Next up was the EU. Notoriously protective of its people, the EU had the most specific and difficult of the regulations to navigate. First there was the EU AI Act. The EU AI Act was published in the Official Journal (OJ) of the European Union on 12 July 2024. The team has created a handy Compliance Checker which helps to classify the system you’re inquiring about with a suggested “risk level.” The risk level then indicates the amount of regulation you can expect to experience when dealing with this system within the EU.
Even this seemingly simple compliance checker necessitated a range of analysis based on my role as an independent researcher and private owner, the modes and methods of operation of the robot, the capabilities of two different models that I was choosing between at the time, the applications I was going to use the robot for vs what it could be used for and so on. It seemed that every time I filled out this checklist, I got a new result. The one consistency was that there seemed to always be an “exclusion” of research and development.
“AI systems and models with the sole purpose of scientific research and development are excluded. For all other systems, research & development activities are likely excluded until your AI system is placed on the market or put into service. Systems and activities that are excluded are not subject to any obligations. For more information see Article 2 points 5a and 5b.”
Next up within the EU was the issue of the GDPR or the General Data Protection Regulation. This was a regulation that I have come up against a lot in my doctoral research on Open Source Intelligence source criticism. It has major implications on data compliance in the EU and created quite an interesting thought experiment for my plans on deployment and experimentation with the robot in “public” settings. There are regulations necessitating encryption, the deletion of private data, data anonymization and more. These are efforts to protect the privacy rights of the citizens of the EU and are fantastic for personal liberty but can present restrictive blocks on the development of new technology.
Although the team has confirmed that no data will be sent back to the headquarters or lab in China, there is still the tricky aspect of identification and facial recognition on my local network. After much discussion, there are two new standards I adopted for the project in order to maintain compliance with GDPR at all times. The first is based on the method of control. The robot will have the capacity for three methods of operation: remote, teleoperation and fully autonomous ai. If the robot is in teleoperation or fully autonomous ai mode, it will passively be scanning using its LiDAR scanner and camera, listening with its microphone array and learning as it moves through the day. If it is in remote mode, I will be controlling its movements and direction and it will rely on my input rather than its own methods and data will not be collected. This will mean that throughout the experiment, when the robot is interacting with the “public” it will always be remote operated so as to maintain compliance with GDPR. When the robot is in my home or my office or a third location, if it is to be operated by teleoperation or autonomously, the people in the room must sign a legal consent form in compliance with GDPR protections to say that their visage and voices can be used in the Robot’s development.
The next EU wide regulation is the European Parliament’s Machinery Directive 2006/42/EC. This directive governs physical Interactions to necessitate incorporating fail-safe mechanisms like emergency stops and collision detection. I inquired with the team about methods of stopping and collision detection and I was able to see videos of lab testing for collision detection performed with the model in constructed environments as well as outside. I learned more about the sensory input of the robot based on its physical hardware including its Intel RealSense D435 Depth Camera, its LIVOX Mid360 3D LiDAR 360 Scanner, its noise and echo cancellation microphone array, and its incredible dexterity within its arms and legs.
Austria
Next to consider is any country-wide regulations. Austria has a project called the AI Mission Austria: AIM AT 2030. This is an interdisciplinary strategy which is the result of a broad stakeholder process in which over 160 experts from various disciplines were involved. It outlines the broader goals of AI research in Austria and maintains an advisory board of experts. As a response to the AI Act which provides for a national body to coordinate and monitor the new regulations, Austria has established the AI Service Desk at Rundfunk und Telekom Regulatory GmbH (RTR).
One of my first local phone calls about the project was to the service desk. As expected, I found that since the advisory board was filled with researchers and experts, they were not in the government office I had rung, but instead dispersed throughout the country in various institutions. I was asked to compose an email that could be forwarded to them and I received a next day response from “The Citizens’ Service of the Austrian Federal Chancellor Karl Nehammer” which exclaimed that they “read about my project with great interest” and advised me to send further information to both the Austrian Federal Chancellery Division VII as well as the Federal Ministry for Climate Action, Environment, Energy, Mobility, Innovation and Technology. I sent outreach to both offices and have yet to hear back!
Vienna
As for the local city perspective, Vienna has the Vienna Smart City Strategy. The main intersections with this project and the strategy are the energy use for the city’s sustainability goals and the public surveillance, consent and data use and anonymization. The latter was addressed on the EU level with the GDPR compliance issues and the project guidelines were already altered in order to bring forth compliance with these goals. The former, however, raises a wonderful point about the energy usage of these humanoid robotics.
There is a twofold analysis of energy usage when using embodied intelligence. One for the body; and one for the intelligence. As for the body, the robot is powered by a 13 string lithium battery. I have purchased two in total, as they each have an approximate two hour run time and I can cycle them to conduct experiments which will necessitate run times longer than 2 hours. The charging of these batteries will impact my apartment energy usage. As such, I have gathered my baseline energy usage in this apartment from my provider Kelag Energie. I will be assessing month to month comparisons between my usage in the two years before the robot to the usage with the robot to get an idea of the increase in energy needed to power it. In addition, the robot, when run on my home wifi or my Bluetooth, will be using far more data than I usually would without it. I will also be tracking this using the baseline of my data usage through Drei Österreich. It is my hope to report these findings to the Vienna Smart City Strategy team for their analysis.
Moving Forward
As the project progresses, I will be updating and cooperating with the relevant authorities and hopefully, inspiring more refined legislation in order to make the transition to a world with domestic humanoids smoother, safer and more advantageous for all involved.
Update as of November 6th:
After much correspondence with the robotics company, I am assured that the G1 EDU is not collecting any face or voice data and that although the video camera and LiDAR sensor are always on, these sensors are used for navigation and obstacle avoidance. I am also assured that with the accompanying app that comes along with the robot, any data including pictures and videos can be directly deleted using the app.
Update as of November 19th:
At this point, I have used email to contact the Office of the Austrian Federal Chancellor Karl Nehammer, the AIM at 2030 Team, the Federal Chancellery Division VII, and the general office of the Federal Ministry for Climate Action, Environment, Energy, Mobility, Innovation and Technology and the Austrian Data Protection Authority. I have not received valuable responses on my emails and followups and as such, I changed my approach and found a valuable resource through calling various offices and asking for recommendations of contacts. One such contact directed me towards speaking with the Austrian Institute of Technology as she believed they would have the direct lines to the relevant regulatory bodies. As of now, I’ve written to them and am awaiting their reply.