Agentic AI and Permission Inheritance Risks: Rethinking Access Control in Autonomous Systems

Home assistant chat-bots, self-driving cars, drones or automated negotiation systems are some of the several examples of autonomous (artificial) agents that have pervaded our society. These agents enable the automation of multiple tasks, saving time and (human) effort. However, their presence in social settings raises the need for a better understanding of their effect on social interactions and how they may be used to enhance cooperation towards the public good, instead of hindering it. To this end, we present an experimental study of human delegation to autonomous agents and hybrid human-agent interactions centered on a non-linear public goods dilemma with uncertain returns in which participants face a collective risk. Our aim is to understand experimentally whether the presence of autonomous agents has a positive or negative impact on social behaviour, equality and cooperation in such a dilemma. Our results show that cooperation and group success increases when participants delegate their actions to an artificial agent that plays on their behalf. Yet, this positive effect is less pronounced when humans interact in hybrid human-agent groups, where we mostly observe that humans in successful hybrid groups make higher contributions earlier in the game. Also, we show that participants wrongly believe that artificial agents will contribute less to the collective effort. In general, our results suggest that delegation to autonomous agents has the potential to work as commitment devices, which prevent both the temptation to deviate to an alternate (less collectively good) course of action, as well as limiting responses based on betrayal aversion.

Hybrid military teams, formed by human warfighters and autonomous artificial agents, represent the technological future of Defence operations. Both the potential and the inherent limitations of current technology are well-known, but the cognitive–behavioral and motivational aspects of human–robot interaction on the battlefield have yet to be systematically investigated. To lay the theoretical and methodological foundation of this scientific investigation, our position paper critically examines how the military personnel’s spontaneous tendency to anthropomorphize artificial autonomous agents can affect operations of hybrid military teams in multiple ways. We will argue that the psychological impact of anthropomorphism on military personnel is neither easily avoidable nor necessarily detrimental. Correctly identifying the multi-level cognitive mechanisms that underpin implicit and explicit forms of anthropomorphism allows us to increase the efficacy of human–agent interaction. We will argue that, within hybrid teams, the capability to communicate with teammates, allies, civilians, and adversaries relies on embodied social cognition processes that are inherently geared toward anthropomorphism and leverage its effects. By updating both the design of autonomous artificial agents and the training of human troops to account for these processes, their reciprocal coordination can be augmented.

What if an AI chatbot could operate a task entirely on its own in software and/or hardware? What if, when you asked your favorite AI chatbot to schedule a meeting with someone, it created your calendar appointment, contacted the person, negotiated the time slot, and sent reminders, handling any situation that arose, entirely without your further input? These so-called "autonomous" or "agentic" AI models are rapidly emerging. In this chapter, we explain how autonomous AI systems already exist now and provide an overview of what they can do. We also discuss the implications this new AI paradigm could have for work and decision-making (Bryson et al., 2020; Acharya et al., 2025; Hughes et al., 2025). We frame agentic AIs as general-purpose computing systems that are able to perform an open-ended set of tasks in various software and hardware environments on behalf of or rather than just in support of human users. Central to our conceptualization of autonomous agents are three properties: the ability to act (and finalize tasks) on behalf of users rather than just in support of users; the ability to execute a wide variety of different tasks that are not limited to query-response or dialog functions; and the ability to work and/or interact in the physical world as well as in virtual environments. We define Agentic AI as those technologies and applications that are based on Large Language Models and their variants that exhibit the properties of such autonomous agents. Commercial "Agentic AI" systems and apps utilize also other AI-centric approaches including multimodal or evolutionary learning and/or hybrid models that integrate LLMs with neural-symbolic, cognitive architecture, or common sense reasoning components. Initial projects in the relatively newly emerging field of Agentic AI focus either on using LLMs to augment existing automation technology or on autonomous process optimization and decision-making. But fully enabling Agentic AI would involve a significant redesign of current processes to fit into an autonomous model that assigns agencies to both humans and agents, a design that allows agents to perform their task autonomously with assistance from human users only when necessary and that allows human users to monitor agent operations and interventions at low cost (Stone et al., 2016; Russell & Norvig, 2020; Sapkota et al., 2025).

Cloud identity management represents a transformative force in telecommunications and finance, revolutionizing how organizations approach security, operations, and customer experiences in increasingly distributed environments. Through innovative technologies including federated identity frameworks, single sign-on architectures, and adaptive authentication mechanisms, cloud identity solutions enable organizations to overcome traditional limitations of fragmented identity infrastructures while addressing industry-specific challenges. In telecommunications, these solutions facilitate optimized customer onboarding, multi-platform service management, and secure access provisioning across network services. Financial institutions leverage cloud identity management to enhance regulatory compliance, prevent fraud, and build customer trust through balanced security and experience considerations. The architectural evolution from traditional perimeter-based approaches to context-aware, attribute-based models demonstrates the strategic significance of identity management as a foundational element of digital transformation rather than merely a security control mechanism. Implementation experiences across both sectors reveal substantial benefits in operational efficiency, security posture, and customer experience, positioning cloud identity management as an essential capability for competitive advantage in rapidly evolving digital ecosystems.

This study investigates the spatial distribution of dengue cases in the city of Rio de Janeiro by employing neighborhood-level partitioned variance analysis, k-means clustering, and multivariate linear regression modeling. By analyzing weekly dengue case reports from the Brazilian Ministry of Health alongside meteorological data from AlertaRio mesonet stations, we reveal a significant correlation between urban thermal structures—specifically the Urban Heat Island (UHI)—and dengue incidence. Notably, increases in daily maximum and minimum temperatures precede peaks in dengue cases by up to one week, indicating a potential influence of urban heat on dengue transmission and severity. The variance analysis identifies four distinct neighborhood clusters, each accounting for a substantial portion of the total variance in dengue incidence, with a confidence level exceeding 90%. These findings highlight the intricate interplay between urban environments, local climate patterns, and the proliferation of vector-borne diseases such as dengue. Furthermore, they underscore how critical socio-economic factors, such as inadequate infrastructure and limited access to healthcare, exacerbate community vulnerability. This study emphasizes the need for integrated urban planning and public health strategies to mitigate health risks associated with UHI, ultimately enhancing resilience against vector-borne diseases in densely populated urban areas.

This study aims to analyze the message of Taesanggameung-pyeon published by the Joseon royal family in the 19th century through text mining. As it was issued by the state, the text of Taesanggameung-pyeon occupies a special position. In addition, the title and genre regulation of the book constitute a superficial message centered on “Taoism, the transcendental, good and evil, and cause and effect.” Superficial messages may trigger cognitive bias in the reading process. Data mining enables a new interpretation of texts that are prone to cognitive biases in that it treats fine-scale text information, which is easy to miss by human attention, as having the same value as information in units that human cognition readily understands. In this article, frequency analysis of vocabulary, visualization of vocabulary characteristics, and topic-oriented visualization were conducted to determine the actual message delivered to the readers of Taesang󰠀gameung-pyeon, which has a strong superficial message. For the linguistic approach and visualization, quantitative analysis techniques such as word cloud, TF-IDF, word2Vec, K-means clustering, and topic modeling were used. Following text mining focused on the vocabulary used in Taesanggameung-pyeon, the ethics emphasized in Taesanggameung- pyeon is not religion-specific, but has general applicability, and the desire that the Gam Eung Story satisfies is a transcendent desire. Rather, it was confirmed that it was a desire to make up for worldly and everyday shortcomings. This affords new insights in that it is distinct from the results of earlier qualitative studies.

The absence of gas stations built in Cibeber Subdistrict is not balanced with the high level of transportation use for ease of mobility among residents. The purpose of this research is to cluster data using K-Means clustering and spatial modeling to provide a potential location for the construction of gas stations in Cibeber District. Based on the research process that has been carried out using RStudio, the potential villages for the construction of gas stations consist of four villages, namely Cikotok, Cibeber, Neglasari, and Wanasari. As for the results of spatial modeling, Cibeber District has a total of 862 potential location points, and within the scope of potential villages, namely four villages, there are 233 potential location points. Then, after being processed with weighted products for optimization and getting the best location results, 3 potential locations were obtained, namely Tegalumbu Village located in Wanasari Village, Nagrak Village located in Cikotok Village, and Cinangga Village located in Cibeber Village.

Artificial autonomous agents and robots interacting in complex environments are required to continually acquire and fine-tune knowledge over sustained periods of time. The ability to learn from continuous streams of information is referred to as lifelong learning and represents a long-standing challenge for neural network models due to catastrophic forgetting in which novel sensory experience interferes with existing representations and leads to abrupt decreases in the performance on previously acquired knowledge. Computational models of lifelong learning typically alleviate catastrophic forgetting in experimental scenarios with given datasets of static images and limited complexity, thereby differing significantly from the conditions artificial agents are exposed to. In more natural settings, sequential information may become progressively available over time and access to previous experience may be restricted. Therefore, specialized neural network mechanisms are required that adapt to novel sequential experience while preventing disruptive interference with existing representations. In this paper, we propose a dual-memory self-organizing architecture for lifelong learning scenarios. The architecture comprises two growing recurrent networks with the complementary tasks of learning object instances (episodic memory) and categories (semantic memory). Both growing networks can expand in response to novel sensory experience: the episodic memory learns fine-grained spatiotemporal representations of object instances in an unsupervised fashion while the semantic memory uses task-relevant signals to regulate structural plasticity levels and develop more compact representations from episodic experience. For the consolidation of knowledge in the absence of external sensory input, the episodic memory periodically replays trajectories of neural reactivations. We evaluate the proposed model on the CORe50 benchmark dataset for continuous object recognition, showing that we significantly outperform current methods of lifelong learning in three different incremental learning scenarios.

ObjectivesTraditional approaches in Alzheimer’s disease (AD) research examine cognitive symptoms in isolation, potentially overlooking dynamic interrelationships among impairment domains. This study employed network analysis to examine structural organization of cognitive and functional domains in mild (mAD) and moderate-to-severe (Mod-sAD) Alzheimer’s disease, aiming to identify stage-specific symptom structures and inform targeted interventions.MethodsA cross-sectional study included 134 participants diagnosed with AD according to DSM-5 criteria. Participants were classified into mAD (n=37) and Mod-sAD (n=97) groups. Regularized partial correlation networks with extended Bayesian information criterion regularization examined symptom interdependencies across six CDR domains: memory, orientation, judgment and problem solving, community affairs, home and hobbies, and personal care. Network comparison tests and centrality analyses identified structural differences between disease stages.ResultsThe Mod-sAD group demonstrated significantly higher impairment scores across all domains (p < 0.001) with large effect sizes (Cohen’s d: 1.83-2.71). Network analysis revealed increased global strength in Mod-sAD versus mAD networks (2.60 vs. 2.49, p < 0.05), indicating greater symptom interconnectedness in advanced stages. Centrality analyses revealed fundamental reorganization: memory emerged as most central in Mod-sAD (strength = 1.62), while judgment and problem-solving showed highest centrality in mAD (strength = 1.65). Orientation centrality increased substantially across progression (strength: -1.32 to 0.40).ConclusionsAD progression features increasing network density and centrality shifts from executive-function-centered networks in mild AD to memory-centered networks in moderate-to-severe stages. Findings suggest stage-specific interventions: executive enhancement in mild AD and memory-focused approaches in advanced stages.

In this paper we presented a framework for action descriptions and its connection to natural language interfaces for artificial agents. The core point of this approach is the use of a generic action/object hierarchy, which allows interpretation of natural language command sentences, issued by a human user, as well as planning and reasoning processes on the conceptual level, and connects to the level of agent executable actions. The linguistic analysis is guided by a case frame representation, which provides a connection to actions (and objects) represented on the conceptual level. Planning processes can be implemented using typical precondition and effect descriptions of actions in the conceptual hierarchy. The level of primitive actions (leaf nodes of this conceptual hierarchy) connects to the agents' executable actions. The level of primitive actions can thus be adapted to different types of physical agents with varying action sets. Further work includes the construction of a suitable, general action ontology, based on standard ontologies like FrameNet (ICSI, 2007), Ontolingua (KSL, 2007), Mikrokosmos (CRL, 1996), Cyc (Cycorp, 2007), or SUMO (Pease, 2007; IEEE SUO WG, 2003), which will be enriched with precondition and effect formulas. Other topics to be pursued relate to communication of mobile physical agents (humans) in a "speech-controlled" environment. The scenario is related to the "smart house" but instead of being adaptive and intelligent, the house (or environment) is supposed to respond to verbal instructions and questions by the human user. A special issue, we want to address, is the development of a sophisticated context model, and the use of contextual information to resolve ambiguities in the verbal input and to detect impossible or unreasonable actions.

The research of autonomous artificial agents that adapt to and survive in changing, possibly hostile environments, has gained momentum in recent years. Many of such agents incorporate mechanisms to learn and acquire new knowledge from its environment, a feature that becomes fundamental to enable the desired adaptation, and account for the challenges that the environment poses. The issue of how to trigger such learning, however, has not been as thoroughly studied as its significance suggest. The solution explored is based on the use of surprise (the reaction to unexpected events), as the mechanism that triggers learning. This thesis introduces a computational model of surprise that enables the robotic learner to experience surprise and start the acquisition of knowledge to explain it. A measure of surprise that combines elements from information and probability theory, is presented. Such measure offers a response to surprising situations faced by the robot, that is proportional to the degree of unexpectedness of such event. The concepts of short- and long-term memory are investigated as factors that influence the resulting surprise. Short-term memory enables the robot to habituate to new, repeated surprises, and to “forget” about old ones, allowing them to become surprising again. Long-term memory contains knowledge that is known a priori or that has been previously learned by the robot. Such knowledge influences the surprise mechanism, by applying a subsumption principle: if the available knowledge is able to explain the surprising event, suppress any trigger of surprise. The computational model of robotic surprise has been successfully applied to the domain of a robotic learner, specifically one that learns by experimentation. A brief introduction to the context of such application is provided, as well as a discussion on related issues like the relationship of the surprise mechanism with other components of the robot conceptual architecture, the challenges presented by the specific learning paradigm used, and other components of the motivational structure of the agent.

Cloud computing offers many benefits to the IT industry by making available the services and resources that helps them to proliferate or decrease their organizational resources automatically on demand. On the other hand, organizations are still uncertain about the security and privacy of their sensitive information (for instance the identity credentials) in the multitenant environment of the Cloud. Many security systems have been devised for the protection of resources in Cloud environments. Identity Management Systems, in this regard, play a vital role in ensuring effective user authentication, provisioning, de-provisioning and access control decisions. Many Cloud IDMSs have been proposed until now claiming to offer flexibility, agility and robustness. However, no comparative analysis of such Cloud based IDMSs has been performed so far, as to the best of our knowledge there exists no specific criteria against which one can evaluate an IDMS on Cloud. This paper proposes an assessment criterion for the evaluation of Cloud based IDMSs, comprising of potential security features that are positively imminent for the assessment of Cloud based IDMSs. Furthermore, analysis of Cloud IDMSs is presented based on the proposed assessment criteria. Potential research directions in the area of Cloud identity management and security are also discussed.

Biological inspiration of animal behavior, nervous systems and natural evolution mechanisms, allow the construction of artificial Autonomous Agents (AAs) that, as animals, could work very well in the real world. This paper uses this inspiration to analyze and simulate evolutionary mechanisms capable of creating and developing different neural network topologies with increased complexity. This complexity is related with the behavior repertoire presented by the network, increasing the survival chances of the agents in a given world. Here is described a system capable of simulating different environments where the AAs operates. Through the use of evolutionary programming techniques, the system generates different artificial neural networks that connect the animats' sensors to its actuators. We introduce and describe a lot of cost functions associated with the neural network complexity, leading to a multi-objective problem, very difficult to solve. Different experiments were made in environments with various levels of complexity. The obtained results confirming the hypothesis in a way that, simpler behaviors can be implemented by feedforward neural networks. As the environment becomes more complex, it is necessary more complex behaviors that only can be implemented by using recurrent neural networks

Automatic discovery and indexing in sign language videos would provide many valuable services such as search and retrieval in deaf social and news media which consists of sign language videos often with no associated texts available. As some of the context words in sign language conversations are fingerspelled, the discovery of fingerspelled letters plays a significant role. In this study, k-means clustering and Gaussian Mixture Model methods are used for the unsupervised discovery of fingerspelled letters. The clusers which mostly contain hand shapes corresponding to transitions between letters are eliminated based on their distortion level and Bayes information criterion values. The performance of the clustering methods are evaluated with the help of Adjusted Mutual Information metric. According to the results obtained with this metric, using a Gaussian Mixture Model for clustering yields better results than k-means clustering.

McLachlan (2011) and Vermunt (2011) each provided thoughtful replies to our original article (Steinley & Brusco, 2011). This response serves to incorporate some of their comments while simultaneously clarifying our position. We argue that greater caution against overparamaterization must be taken when assuming that clusters are highly elliptical in nature. Specifically, users of mixture model clustering techniques should be wary of overreliance on fit indices, and the importance of cross-validation is highlighted. Additionally, we note that K-means clustering is part of a larger family of discrete partitioning algorithms, many of which are designed to solve problems identical to those for which mixture modeling approaches are often touted.