The technology environment is expanding beyond large language models (LLMs) and software-based artificial intelligence agents into embodied AI—systems designed to move and interact within the physical world. As organizations and nations invest heavily in intelligent robotics, ensuring the security and integrity of these systems has become a priority. Threat actors are already exploring ways to compromise these emerging technologies to gain strategic advantages.
Industry experts observe that embodied AI systems may drive a major industrial shift, following historical advances in manufacturing, electricity, and telecommunications.
Joseph Rooke of Recorded Future will address these dynamics at the upcoming Infosecurity Europe conference. His presentation covers power politics, supply chain risks, and security incident scenarios involving embodied AI systems. Rooke notes that while the industry is advancing quickly, the current security posture of many of these systems requires immediate attention.
Security considerations for intelligent robotics
As robotic kinematics advances alongside other artificial intelligence technologies, the deployment of embodied systems is expected to scale rapidly. Morgan Stanley projected that China alone could deploy approximately 300 million embodied AI units by 2050 across industrial and general use cases. The integration of humanoid robots is already a component of Chinese state initiatives, and drone systems have seen notable use in the Russia-Ukraine conflict. Commercially, robotic systems are increasingly accessible to the public.
Embodied AI introduces unique physical and digital safety considerations. Systems operating in homes and industrial environments continuously process data through visual and audio sensors. If unauthorized access occurs, these data streams can be exposed. Furthermore, the physical interfaces themselves present new pathways for system compromise.
Security researchers have conducted extensive evaluations of popular models, such as Unitree humanoids. These assessments revealed that the devices transmitted user data to servers in China without explicit consent. Furthermore, researchers identified undocumented access mechanisms that permitted full, unauthorized control of the systems. These vulnerabilities could be compromised rapidly and possessed the capability to spread wirelessly between units. As Rooke observes regarding the findings, "That really was terrifying—iRobot sort of stuff. A fleet could be compromised in this case."
There is a recognized pattern of bringing AI technologies to market before fully addressing their security models. Rooke emphasizes that in the rapid development of the US robotics market, organizations must prioritize comprehensive risk assessments to avoid overlooking critical vulnerabilities.
Supply chain security and the mining sector
Securing intelligent robotics extends well beyond the physical units. The supply chains required to build embodied AI systems include AI models, data centers, semiconductors, and energy infrastructure. These maturing supply chains are increasingly targeted by state-sponsored threat actors conducting commercial and geopolitical espionage.
Rooke draws particular attention to security incidents within the global mining industry. The production of embodied AI components relies heavily on rare earth elements and critical minerals. Control over these natural resources offers significant market influence, and many valuable deposits are situated in regions with complex regulatory frameworks, including the Arctic and outer space.
Threat actors frequently target this sector to gather intelligence on competitive strategies, disrupt operations, or acquire proprietary mining technologies. Recorded Future has documented several espionage campaigns directed at mining and related industries over the past five years, often involving advanced persistent threat (APT) groups linked to China:
In 2021, APT15 (also known as Nickel or Nylon Typhoon) targeted a Canadian mining company.
In 2025, during a period when China entered seabed exploration partnerships with several nations, multiple APT groups compromised an organization responsible for monitoring and regulating seabed mining.
Between 2021 and 2026, threat actors targeted public and private sector entities in Indonesia, a region where China maintains extensive contracts and natural resource interests, particularly in nickel.
In 2025, YoroTrooper (also known as Silent Lynx), a group researchers assess is based in Kazakhstan, directed operations against Russia's energy, manufacturing, and mining sectors.
During competitive bidding for mining contracts, threat actors may establish unauthorized access within government or ministerial networks to gather reconnaissance. "How can we undercut this bid? It's all about getting that advantage," Rooke says. However, he notes that these current operations are primarily focused on espionage and intelligence gathering rather than destructive activities.
About the author
Nate Nelson is a journalist and award-winning scriptwriter who contributes to Dark Reading and Darknet Diaries. He began his career ghostwriting executive opinion pieces for Forbes and CNBC before transitioning to security journalism at Threatpost, where he covered industry news and trends. He also co-created the technology podcast Malicious Life. Nelson holds degrees from New York University and Bard College, and as a native New Yorker, he maintains the city's traditional confidence while keeping his interactions polite.