Aziz Shuaib Ausi logoAziz Shuaib Ausi
Knowledge Repository

Library

Completed institutional publications — frameworks, reports, toolkits, guides and journal articles, held permanently and available to download.

A refined private library with walnut shelves and warm reading light

Publications

21 publications available to download.

Browse collections Reading journal
Executive GuidePublic 1 min read

Toward a Threat Actor Profiling Taxonomy for Pre-Release Risk Management of Open-Weight Frontier Models

A new research paper identifies a critical gap in pre-release risk management for frontier AI models: the lack of explicit, consistent, and grounded characterization of potential threat actors. It proposes that defining adversaries is essential for effective and comparable risk evaluations. To address this, the paper introduces a six-attribute taxonomy for profiling threat actors, drawing on established literature from terrorism, biosecurity, and cybersecurity.

Published
27 Aug 2026
Version
v1.0
Topic
Ai
Open publication
Executive GuidePublic 1 min read

GGSS: Geodesic-Gated Spherical Steering for Inference-Time Debiasing of Generative Vision-Language Models

Research identifies that generative vision-language models (VLMs) can produce demographically biased outputs, even when visual inputs vary only in controlled attributes like perceived race or gender. Existing debiasing methods are often inadequate for these generative models. A new intervention, Geodesic-Gated Spherical Steering (GGSS), is proposed to address this by steering visual tokens along geodesic arcs within a counterfactual bias subspace, with an adaptive gate to focus corrections on strong demographic signals.

Published
27 Aug 2026
Version
v1.0
Topic
Ai
Open publication
Executive GuidePublic 1 min read

Assessment practices resistant to inappropriate AI use in Open Distance eLearning: insights from Zimbabwe Open University

The integration of artificial intelligence (AI) in education, particularly in Open Distance eLearning (ODeL) environments, has prompted concerns among educators regarding the integrity of traditional assessment methods. This issue is highlighted by students' inappropriate use of AI for academic tasks at institutions like the Zimbabwe Open University. Research efforts are underway to identify assessment practices that foster authentic learning and mitigate the misuse of AI in these settings.

Published
27 Aug 2026
Version
v1.0
Topic
Publisher
Open publication
Executive GuidePublic 1 min read

Computational Orientalism: Measuring Structural Discourse Bias in Large Language Models Using the Middle East Cultural Sensitivity Score (MECSS)

Research indicates that Large Language Models (LLMs) may exhibit structural discourse bias, specifically 'Orientalism,' when representing non-Western cultures, such as the Middle East. This bias stems from reliance on predominantly Western and English-language training data, which can lead to representations that deny agency, prioritize Western frameworks as neutral, and explain regions through externally imposed categories. Traditional fairness metrics are insufficient to detect this nuanced structural bias.

Published
20 Aug 2026
Version
v1.0
Topic
Ai
Open publication
Executive GuidePublic 1 min read

Longitudinal Relational Publics and their Discursive Overlap with Issue Publics

This research introduces a new conceptual framework, the 'longitudinal relational networked public,' to analyze how political discourse emerges and propagates within online spaces not explicitly dedicated to political discussion. It highlights the importance of understanding the dynamics of participants and listeners in these 'online third spaces' over time to determine when and by whom political issues are introduced into nominally apolitical contexts.

Published
20 Aug 2026
Version
v1.0
Topic
Ai
Open publication
Executive GuidePublic 1 min read

One Gate Is Not Enough: Composing Stateful Pre-Action Controls for Agentic AI

This research introduces a formal framework for understanding and managing pre-action controls in agentic AI systems, specifically addressing situations where multiple controls (e.g., authority, resource, evidence gates) govern a single action. The core finding is the concept of 'remediation-induced control coupling,' where the remediation by one control can inadvertently alter the conditions for another, potentially invalidating prior judgments. The paper proposes a protocol to restore per-action soundness in these complex interactions.

Published
20 Aug 2026
Version
v1.0
Topic
Ai
Open publication
Executive GuidePublic 1 min read

Artifact-centered Claim-aware Observability for Autonomous Scientific Agents

The increasing deployment of autonomous scientific agents, which handle tasks from ideation to paper drafting, necessitates advanced observability and auditing mechanisms. Current logging, tracing, and provenance tools are insufficient because failures in these systems are often distributed across multiple artifacts and claims, requiring a more integrated approach to inspection.

Published
20 Aug 2026
Version
v1.0
Topic
Ai
Open publication
Executive GuidePublic 1 min read

Qualified Cross-References as a Verification Method: The Normative Environment of the EU AI Act

Research has introduced a novel model and construction protocol for 'qualified cross-references' in legal texts, specifically applied to the European Union AI Act and its surrounding normative environment. This method aims to improve the clarity, consistency, and verifiability of legal interactions by detailing the nature of links between legal instruments and provisions, moving beyond simple existence of a link to encompass its character, supporting provisions, conditions, and consistency.

Published
20 Aug 2026
Version
v1.0
Topic
Ai
Open publication
Executive GuidePublic 1 min read

What Can Artificial Intelligence Learn from Medicine? Generative Analogies and Reliable Machine Learning Systems

The application of machine learning (ML) in medicine has seen success, yet its foundational epistemic and methodological warrants remain uncertain. Research proposes a generative analogy between clinical translation processes and ML system development, suggesting that the established standards from medical practice could inform and improve the reliability and trustworthiness of ML systems.

Published
20 Aug 2026
Version
v1.0
Topic
Ai
Open publication
Executive GuidePublic 1 min read

The Fabricated Front: Generative AI and the Opacity of Workplace Performance

Recent research from arXiv introduces the concept of 'effort opacity' in workplaces due to Generative AI (GenAI). While much focus has been on productivity and job displacement, this study highlights GenAI's impact on interactional dynamics. It suggests that GenAI systematically decouples observable output from human engagement, weakening reciprocal exchanges and collaborative trust by making interactional cues less diagnostic. This reconfigures workplace interactions, potentially affecting organizational cohesion.

Published
20 Aug 2026
Version
v1.0
Topic
Ai
Open publication
Executive GuidePublic 1 min read

Global Index on Responsible AI 2026 : Conceptual Framework and Methodology

The Global Index on Responsible AI (GIRAI), 2nd Edition, outlines a refined conceptual framework and methodology for assessing responsible AI governance. This updated edition distinguishes between the existence of AI frameworks and their practical implementation, expanding its structure to five thematic dimensions and introducing more detailed variables for evaluating framework quality. The methodology underwent an independent statistical pre-audit to ensure its robustness.

Published
20 Aug 2026
Version
v1.0
Topic
Ai
Open publication
Executive GuidePublic 1 min read

FairGlucose: A CGM Fairness Benchmark Reveals Subgroup Disparities Hidden in Population-Level Validation

Research on CGM-based AI tools highlights a critical issue where population-level accuracy metrics can mask significant performance disparities across diverse patient subgroups. A new benchmark, FairGlucose, demonstrated that while aggregate external validation appears stable, specific demographic strata experience notable variations in forecasting accuracy, particularly for Type 1 diabetes patients.

Published
20 Aug 2026
Version
v1.0
Topic
Ai
Open publication
Executive GuidePublic 1 min read

CentaurBench: Benchmarking LLM Capabilities on Augmenting vs. Automating Real-World Work Tasks

A new research framework, CentaurBench, has been introduced to evaluate Large Language Models (LLMs) based on their capacity to both automate tasks and augment the performance of other agents. Unlike traditional benchmarks that focus solely on automation, this framework assesses how LLMs enhance the output of a lower-capacity worker model across seven real-world tasks, alongside their direct automation capabilities. This shift in evaluation considers the practical application of LLMs as assistants, providing a more nuanced understanding of their utility in collaborative work environments.

Published
20 Aug 2026
Version
v1.0
Topic
Ai
Open publication
Executive GuidePublic 1 min read

Capability-Based Planning for AI Crisis Preparedness

Current government approaches to AI crisis preparedness are hampered by a 'predict-then-act' paradigm that struggles with the inherent unpredictability of AI. Experts and official reviews acknowledge that traditional likelihood-based risk assessments are unsuitable for AI-related risks. A new methodology is proposed, drawing on principles of decision-making under deep uncertainty, which focuses on capability-based planning rather than predictive risk ranking.

Published
20 Aug 2026
Version
v1.0
Topic
Ai
Open publication
Executive GuidePublic 1 min read

Guidance: Academies chart of accounts

The UK Department for Education has issued guidance on the 'Academies chart of accounts,' which details the standardized financial data structure and associated guides for academy trusts to submit their financial returns. This initiative aims to streamline and standardize financial reporting across the academy sector.

Published
20 Aug 2026
Version
v1.0
Topic
Policy
Open publication
Executive GuidePublic 1 min read

Academies financial returns

The UK Department for Education (DfE) has issued guidance regarding the mandatory financial returns that academy trusts are required to submit. The details provided pertain to the specifics of these reporting obligations, indicating an ongoing focus on financial oversight within the education sector.

Published
20 Aug 2026
Version
v1.0
Topic
Policy
Open publication
Executive GuidePublic 1 min read

Decolonizing university education—the concerns consuming Indian students regarding the compulsory isiZulu module at UKZN

A study conducted at the University of KwaZulu-Natal (UKZN) investigated Indian students' attitudes towards a compulsory isiZulu module, which has been in place for approximately a decade. Utilizing both quantitative questionnaires and qualitative interviews, the research assessed student perspectives, the impact on isiZulu usage, and communication abilities post-module completion. The study indicates that social variables contribute to divisions between Indian and Black African communities, and observed limited multilingualism among students.

Published
20 Aug 2026
Version
v1.0
Topic
Publisher
Open publication
Executive GuidePublic 1 min read

Start the semester with one year of Gemini, on us

Google for Education is offering college students globally a complimentary 12-month subscription to a Google AI plan, specifically Gemini, to commence the academic semester. This initiative aims to integrate advanced AI tools into the educational environment.

Published
20 Aug 2026
Version
v1.0
Topic
Edtech
Open publication
Executive GuidePublic 1 min read

Runtime Governance for Agentic AI: Action-Boundary Control with Trusted Provenance and Fail-Closed Execution

The advent of agentic AI systems introduces a new class of safety challenges, shifting from harmful content generation to harmful operational side effects caused by AI-initiated actions such as file modifications or workflow changes. Traditional prompt-level governance is insufficient to contain these risks as it lacks an execution boundary. A proposed solution, Aegis, addresses this by implementing a runtime governance system that interposes a trusted decision layer between the AI model's action proposals and their execution. This system evaluates proposals against established policies, ensures trusted provenance, operates on a fail-closed principle, and can route complex decisions for quorum-based authorization.

Published
20 Aug 2026
Version
v1.0
Topic
Ai
Open publication
Executive GuidePublic 1 min read

Towards welfare-oriented recommendations in activity-travel behavior

Current recommender systems (RS) in activity-travel behavior often fail to adequately account for user welfare, potentially leading to recommendations that leave users worse off than alternative choices. This research introduces a framework to address this gap by focusing on 'net utility' to ensure recommendations actively improve user welfare, particularly relevant where users incur non-recoupable costs like time and energy.

Published
20 Aug 2026
Version
v1.0
Topic
Ai
Open publication
Executive GuidePublic 1 min read

Traceable Trust for action-ready artificial intelligence in bioscience

Artificial intelligence (AI) is increasingly integrated into bioscience operations, performing tasks from predicting structures to optimizing experiments. This research highlights the critical juncture when AI outputs are used to guide laboratory action, emphasizing the need for a trustworthy and reviewable process. It introduces 'Traceable Trust,' a framework designed to assess and manage this output-to-action boundary through structured inquiry.

Published
20 Aug 2026
Version
v1.0
Topic
Ai
Open publication