Editorial: Talking fabrics – a new era in autonomous AI textile wearable (AITW) technology

In a world where technology constantly transforms the boundaries of possibility, we find ourselves at the intersection of a quiet but powerful revolution: the fusion of artificial intelligence (AI) with the fabrics that we wear as garments. This remarkable evolution in wearable technology is not simply a step forward but a giant leap into an era where our clothing becomes an extension of our own cognitive experience. In fact, it provides what AI needs to go from “chatting unaware of the user” to becoming more intelligent by understanding the user's physical and environmental conditions and hence providing “contextual understanding”, which is the next most important requirement in AI's quest to achieve large action models (LAM) for higher intelligence. The research presented here introduces this new frontier of “AI textiles (AIT).”The notion of intelligent textiles isn't new. Increasingly, fabrics are considered organic beings, capable of understanding and reacting to human conditions and environment (Stylios, 2020). What makes this work by Aliyana and Stylios exciting is its tangible manifestation of that vision. AI wearables (AITW) is a prime example of the realm of autonomous cognitive ability. An AIT is a system that listens, monitors and measures processes and converses with its wearer. Powered by self-harvested energy through triboelectric nanogenerators (TENGs) embedded in the textile, an AIT monitors biomechanical data (movement, temperature, breathing, heart rate, resistivity, touch, etc.), turns that data into contextual information so that the communication between the BOT and the user is moving to a higher level of AI than at present. By embedding AI in a fabric that responds to real-time data, AIT transcends the traditional role of wearables. It actively participates in all activities of its wearer, offering personalized feedback, from gait analysis through voice interaction and is responsive to the user's real-world needs. We can call these Autonomous Intelligent Agents (AIAs) or Autonomous Intelligent Copilots (AICs). They understand not just what the user says, but also the context, intent and situational background. They personalize responses and actions based on user history, preferences and real-time data. Make autonomous decisions and take proactive steps to help or alert the user and continuously learn and adapt, becoming more effective over time, rather than simply reacting to commands, as bots do at present. These autonomous agents will be the textile or the wearable garment itself, hence calling them AITAs or AITWAs, respectively.The evolution from passive question-answering AI like ChatGPT to advanced AAI represents a significant leap toward higher intelligence. This shift transforms AI from a reactive tool into an autonomous, proactive system capable of complex reasoning, decision-making and real-world action. Let us break down the reasons why and how this advancement occurs. This is a leap in intelligence because AIA surpasses models like ChatGPT in three critical dimensions: firstly, in autonomy and proactivity; unlike ChatGPT, which waits for user prompts, AIA independently initiates actions, sets goals and adapts strategies based on environmental data (e.g. user habits, sensor inputs, etc). An example might be an AIA monitoring a user's health metrics might proactively adjust room temperature or suggest rest without being asked. Secondly, in contextual depth and memory, ChatGPT resets context between interactions, while AIAs maintain long-term memory, retaining user preferences, historical data and environmental states to inform decisions. This enables nuanced personalization, such as refining recommendations based on past behaviour. Thirdly, in task execution and integration, AIAs integrate with external systems (APIs and IoT devices) to execute multi-step workflows like booking travel or controlling smart homes far beyond ChatGPT's text-based responses. They solve complex problems by analysing real-time data (e.g. optimizing logistics routes) (Constantino, 2025; Nahavandi et al., 2022).This advancement is achieved because AIA leverages four technical innovations to enable higher intelligence:Therefore, as we can see, AITs mark a paradigm shift by merging contextual awareness, autonomous decision-making and real-world action, transforming AI from a conversational partner into an intelligent collaborator. This is achieved through architectures that emulate human-like reasoning, continuous learning and seamless system integration, enabling agents to tackle dynamic, multifaceted challenges that passive models cannot. Table 1 shows a comparison between a chatbot like ChatGPT and AIT agents.Networked AITA/AITW agents' communication is where agents share contextual data from textile wearables about users during interactions, which could transform how humans connect by making conversations more informed, empathetic and efficient. This system would function like an advanced social layer over platforms (e.g. Facebook, X and Instagram), but with explicit user consent and real-time data exchange between AIAs, see Figure 1. Below, we analyse its potential to enhance interactions, its broader effects and key regulatory considerations.As we have discussed, textiles provide the ability of AIA to have continuous contextual enrichment and become AIA themselves as AITAs. AITAs could share non-sensitive, user-approved context (e.g. “Alex just returned from Japan” or “Sam prefers video calls over texts”) before or during conversations. This allows participants to tailor discussions to shared interests or recent events, fostering deeper engagement. Textiles worn by users can provide AI agents the ability of “proactive assistance”; agents might alert users to “potential sensitivities” (e.g. “avoid discussing travel: Maya's flight was cancelled today”) or suggest “conversation starters” based on mutual hobbies, reducing social friction and building rapport textiles will also provide “efficiency in networking”; for professional settings, agents could exchange role-specific backgrounds (e.g. “Taylor's expertise: IoT security”), enabling more focused and productive collaborations. In fact, these AI agents are the AITAs/AITGAs themselves.It is inevitable that there are incredibly positive effects with this new technical breakthrough but also negative effects, which will need mitigation. A welcome positive effect is the deeper personalization, in which conversations feel more relevant and empathetic, as agents provide real-time context about mood, preferences or milestones (South et al., 2025; IBM, 2024). But privacy could be eroded by oversharing without granular consent, exposing sensitive traits (e.g. health status) or enabling manipulation. Another welcome aspect is the build-up of stronger relationships, because shared context fosters trust and common ground, mimicking the intuition of close friendships, but at the same time, users may feel social pressure and might alter behaviour if aware they're being “analysed,” leading to inauthentic interactions, which may be a negative effect. The technology can have another positive effect in conflict reduction because agents could flag incompatible communication styles (e.g. “Jamal dislikes interruptions”) to prevent misunderstandings (South et al., 2025). However, this feature also risks amplifying bias if agents rely on incomplete data, reinforce stereotypes or inherit algorithmic biases that may influence user perceptions.As textile wearable garments can provide contextual-based data from the wearer to enable the AI agents to become more intelligent, personal and effective, as we have seen, several regulatory and ethical constraints must be put in place; for example, the wearer should have a “consent-first design” where the user must explicitly opt into data sharing per interaction, with options to limit context (e.g. “share my hobbies but not my location”). Regulations like GDPR and the EU AI Act mandate this granular control. Then the question of “transparency”, in which agents should disclose when/why context is shared (e.g. “Based on your LinkedIn, I suggested your startup experience”) to maintain trust. And very importantly, the insurance of “human oversight”, in which high-stakes scenarios (e.g. conflict mediation) should require human review to override agent suggestions, aligns with GDPR's Article 2 “right to explanation” (European Union, 2016; UK Government, 2018). In consequence, AI agent/textile wearable networking could make human interactions more intuitive and meaningful by leveraging contextual data, turning superficial chats into enriched exchanges. However, this hinges on strict consent protocols, transparent data usage and bias mitigation to avoid privacy harms or social manipulation. While regulation (like the EU AI Act) is evolving to address these risks, the priority remains for user autonomy: ensuring agents enhance, not dictate, human connection (European Union, 2024).The future will involve textile wearables carrying AI-enabled devices with nano chips for local processing while leveraging larger cloud-based models, creating a hybrid architecture optimized for performance, privacy and efficiency. This approach aligns with the evolution of AIA (specifically edge AI agents), which combine autonomy, contextual awareness and distributed intelligence. In this case, AIGA became Edge AIGAs. These wearables are already autonomous in terms of energy and sensing by virtue of their yarns and fabric structures, and ideally, it will be their battery-free provision to be maintained so that their energy harvesting ability will be sufficient to make the AI nano chips for local processing work battery-free. Below is a breakdown of how this works:Let us discuss how this can work in a real end-use scenario.So, this is a real scenario which benefits from a textile enabling contextual understanding by sensing physiological and environmental data from the wearer and so lifting the intelligence of the agent to edge, which stays locally in the wearable textile system, facilitated by a nano chip in the textile and the cloud-based agent as seen in Table 2.There is an interesting and more sustainable AI industry in a future outlook based on the above considerations, such as 5 G/IoT Synergy, where there will be a faster edge-cloud synchronization for complex tasks (e.g. AR-assisted surgery) (MarketsandMarkets, 2024). Then, self-learning wearables (SLW) embedded with nano chips will locally retrain models using federated learning, reducing cloud dependence (Lyzr, 2024). This will make wearable textile garments truly autonomous edge agents on their own right, increasing the need for better implementation of ethical guardrails by easier-to-administer user-controlled consent layers (e.g. “Share heart data only for emergency alerts”) (MIT Computing, 2024; MarketsandMarkets, 2024). Therefore, carrying AI-enabled wearables with nano chips is essential for real-time responsiveness and privacy, while cloud-based models handle deeper intelligence. AI agents bridge these layers, creating a cohesive, efficient system where edge devices act as autonomous “first responders” and the cloud serves as a “strategic brain.” This architecture will dominate future tech and will change our lives forever. It is expected to replace Google, Facebook, Yahoo, X and Instagram. To achieve an advanced AIA capable of context awareness, personalization, decision-making and proactive assistance, the hardware and software must integrate edge-device efficiency with cloud-scale intelligence. Below is a summary of requirements based on current technological advancements:Edge textile wearables/IoT will need ultra-low-power processors, i.e., nano-chips like NeuroSense (consuming <100 µW) for real-time sensor data processing (e.g. motion), especially since their power (TENG) will have to be provided by the movement generated in the textile by the wearer (Aliyana et al., 2024). Sensing will be inherent to the textile-based wearable, and it will depend on the type of fibres the textile is made of. Finally, connectivity will be based on Bluetooth Low Energy (BLE), or 5G RedCap/NB-IoT will be employed to ensure minimal energy consumption during cloud communication (NeuroSense, 2024). In contrast, cloud infrastructure will rely on high-performance accelerators, such as NVIDIA H200 SXM, H100 SXM, TPUs and FPGAs, for training large AI models and executing complex inference tasks. In contrast, the cloud infrastructure will rely on high-performance accelerators, such as GPUs (NVIDIA H200 SXM and H1oo SXM, TPUs and FPGAs), for training large AI models and complex inference tasks. These chips consume hundreds of watts of power and require robust power delivery and thermal management systems. Scalable, high-throughput storage using NVMe SSDs will also be essential for rapid access to large datasets during training and retrieval. While edge devices focus on energy-efficient sensing and minimal computation, cloud systems are optimized for performance, throughput and large-scale model hosting, reinforcing the distributed nature of modern AI-powered wearables.Edge Software, which is based on the textile wearable agent, stays local within the nano chip-based system by means of tiny ML frameworks, such as TensorFlow Lite or Edge Impulse, for deploying lightweight models on wearables (Latent AI, 2024). The need for a real-time OS, optimized for low-latency textile wearable sensing data processing and local decision-making. And importantly, federated learning enables on-device model updates without raw data leaving the device (Latent AI, 2024; Zilliz, 2024). Cloud software must perform AI orchestration so that platforms like SmythOS or Latent AI will be used to manage agent workflows, versioning and scaling (Milvus, 2024). MLOps tools will be used for continuous deployment, monitoring and updating of edge models (e.g. latent AI's telemetry) (Zilliz, 2024). And finally, facilitate APIs/services for cloud-based AI APIs (e.g., Azure Cognitive Services) for complex tasks like NLP. So, if we were to look at this hybrid architecture workflow, it would consist of the following layers:For this hybrid architecture workflow, we will need the following development requirements: hardware with ≥8 GB RAM, multi-core processors (Intel i5+/AMD Ryzen) and 20 GB + storage for testing wearable apps. And tools like SDKs for target platforms (Android/iOS), BLE debuggers and AI optimization suites. In essence, the future demands capable, durable and comfortable textile wearables; hardware that balances ultra-low-power edge processing with cloud-scale computation, coupled with software that enables seamless, privacy-preserving intelligence across this hybrid architecture (Nahavandi et al., 2022; Shajari et al., 2023; EEJournal, 2024; Google Cloud, 2024). And preferably everything is done without reliance on traditional batteries, drawing power directly from an array of TENG-integrated textiles. Based on this architecture workflow, a high-level sketch of system components and data flow is provided in Figure 2, with each key element described in the subsequent sections.Key components explained:Data flow sequence:This architecture balances low-latency edge processing with cloud-scale intelligence, enabling responsive, context-aware AI agents while maintaining data privacy and energy efficiency. The visual diagram in Figure 2 illustrates these interactions in greater detail.We have implemented and validated this concept through the development of autonomous, battery-free sensing textiles capable of capturing the range of user data – movement, temperature, breathing, pressure, resistivity, touch intensity, heart rate, EKG and ECG – depending on the choice of the functional textile material used. The fabric is designed into a wearable vest and leggings, which, through a WebSocket, transmit the data to ChatGPT's LLM, where data are trained, and a meaningful conversation is then performed between the user and ChatGPT. The T-TENGs system demonstrated groundbreaking performance, with triboelectric outputs peaking at 800 V and 80 μA under maximum frequency and load conditions corresponding to a power density exceeding 3 W/m2, among the highest ever recorded in textile-integrated TENG systems, Figure 3. Each footstep generated up to 0.35 μC of charge during high-intensity activity. With frequency responsiveness optimized between 2 and 8 Hz and signal stability maintained over 1,000 cycles, the architecture ensures robustness and repeatability. Furthermore, textile bot's step-recognition capabilities were validated with industry-leading precision, demonstrating 95.3% accuracy during wearable use, underscoring the consistency and robustness of the embedded sensing architecture. The false positive detection rate remained at while real-time data from the textile to the an of We then the wearable and the but this were as in Figure The conversation became more meaningful and personal and and the more learning, the more effective it A of this intelligence is its with which transforms raw biomechanical data into dynamic, personalized conversational Each data step gait and charge is into and by the to health or of wearable technology focused on and the communication of data by wearables on cloud-based or external devices to the What AIT introduces is an new in which the garment is an system capable of with the This approach is truly because it autonomy, batteries, real-time sensing data and into a wearable These textiles are through energy from the user's and by virtue of their perform the need for external or power of embedded in the textile means that energy is into the making it and (Aliyana and 2023; Aliyana et al., 2025; et al., 2025). This system ensures the textile remains and autonomous, offering users an incredibly efficient and experience. The of AI, through the AI to a new The ability to in provide contextual understanding and learning, and it from AI to more exciting about this work is the depth of that into the of material and wearable and energy real-time sensing and AIT represents a of these enabling a truly seamless experience. The of is an example of how is the potential for wearables. These biomechanical a breakthrough in technology that new for autonomous wearable systems. The textile yarns and the between and The research how material and garment can for energy harvesting while ensuring during This is important for wearable technology that needs to integrate into without user Furthermore, the of AI with this fabric allows for a interaction between the wearer and their the for personalized AI systems that with the user's needs. What truly sets AIT is its ability to provide The real-time processing and of data the system to responses to the user's as go about their may but the material by humans – textile – is to enable a significant leap in AI This could AI from being a question-answering chatbot to a more intelligent system capable of contextual human-like intelligence. In this with the and we have and the hardware and software requirements for such We also demonstrated how current in can make this a the work of and Stylios as to this we new a TENG fabric can function as an AI textile agent with an an autonomous system that and The fabric provides energy-efficient sensing while the the intelligence for autonomous decision-making and action. This enables across and smart AI agent such as autonomy, learning and Furthermore, a TENG fabric with an AI can be as an Edge This system the autonomous, sensing of TENG textiles with AI processing to an intelligent, wearable a TENG fabric with an AI an Edge energy harvesting and sensing with AI such as personalization, and networking and This transforms passive garments into systems with The can be with to wearable garments. By an Edge we bots from passive into intelligent This fusion of physical edge processing and conversational AI enables interactions based on real-time and environmental data, proactive such as health or and seamless in this the for a networking In this we have demonstrated how this approach a paradigm where current bots from reactive tools to context-aware as capable of and in