Tue 16 June 2026
Daily Brief — Curated and contextualised by Best Practice AI
Anthropic Negotiates, OpenAI Spends Billions, and Nvidia Tests Investor Nerves
TL;DRAnthropic is in talks with the US to lift restrictions on its AI models. OpenAI spent $34 billion last year, preparing for an IPO. Nvidia plans to raise over $25 billion in its first bond deal since 2021, testing investor appetite amid AI sector borrowing. AI is driving productivity and job growth, but creating a two-track labor market. US policies have inadvertently boosted China's open AI ecosystems.
The stories that matter most
Selected and contextualised by the Best Practice AI team
Cognitive Debt: AI as Intellectual Leverage and the Dynamics of Systemic Fragility
arXiv:2606.15078v1 Announce Type: new Abstract: We develop a formal theory of cognitive debt: the stock of unverified reasoning obligations that accumulates when individuals use AI as a substitute rather than a complement for first-principles cognition. The model features two state variables per agent, cognitive capital and cognitive debt, and a multiplicative production technology in which cognitive capital functions as collateral that determines the return to AI adoption. We establish six propositions. Rational agents incur positive cognitive debt because the costs are deferred, partially external, and masked by short-run productivity gains. Tranquil periods lower subjective risk assessments, raise AI substitution intensity, and compound leverage, generating a cognitive Minsky moment in which subjective risk falls while true systemic fragility rises. Expected crisis losses are convex in aggregate leverage. Post-crisis, output-target pressure can produce a false-correction loop in which agents patch AI failures with more AI. The decentralised equilibrium over-adopts substitutive AI relative to the social optimum because of systemic risk, cognitive public goods, and arms-race externalities. In a two-type heterogeneous-agent economy, high-cognitive-capital agents adopt AI more intensively and may eventually erode their unaided cognitive capital below that of initially lower-skilled agents.
Trust Between AI Agents: Measuring Formation, Breakage, and Recovery, with Implications for Governing Multi-Agent Systems
arXiv:2606.14923v1 Announce Type: new Abstract: As language-model agents increasingly work in teams, each agent must decide how much to trust its teammates. Yet we lack a standard way to measure trust between AI agents. We propose a behavioral measure based on costly verification. In a cooperative survival game, checking a teammate's work consumes resources, while trusting a wrong answer can be fatal. Relative to a memoryless version of the same model, reduced verification provides an observable measure of trust. Using this framework, we study trust formation, breakage, and recovery across six frontier model snapshots. When paired with a consistently reliable teammate, four snapshots (Claude Opus 4.6, Claude Sonnet 4.6, GPT-5.1, and Gemini 3.1 Pro) reduce verification by roughly 60-85%, whereas two smaller snapshots show little or no such adjustment. Failures reverse this discount, but models differ in how they respond. Some concentrate renewed scrutiny on the culprit, while others become more cautious toward the entire team. Recovery is slower than formation, and clustered failures sustain suspicion far longer than the same number of failures spread apart. These differences have practical consequences. Models that form trust verify less, decide more quickly, and achieve higher payoffs in our environment. By contrast, persistent over-verification is associated with indecision rather than safety. Our results show that trust dispositions can be measured before deployment and suggest that calibration, rather than maximal suspicion, should be the central concern in the governance of multi-agent AI systems.
Economics & Markets
STMicro Eyes $1.5 Billion Convertible Bonds After AI-Fueled Jump
STMicroelectronics NV is looking to raise $1.5 billion from selling debt that can be exchanged into equity after the chipmaker’s shares tripled in value so far this year.
AI Re-Rating Fuels 550% Rally in Hong Kong’s Kingboard Laminates
Kingboard Laminates Holdings Ltd. has rallied more than sixfold this year as investors bet the Chinese supplier to the printed circuit board industry will emerge as a key beneficiary of the AI buildout.
AI strategy affects business valuation: Chicago Booth - Crain's Chicago Business
Investors will put a premium on companies that use AI to deepen customer relationships, improve operations and turn data into an asset.
Follow The Cluster? How To Thrive In The AI Economy
Here is how to understand the co-location premium in the AI economy
Smiles in Profiles: Improving Efficiency While Reducing Disparities in Online Marketplaces
arXiv:2209.01235v5 Announce Type: replace Abstract: Online platforms often have conflicting goals: they face tradeoffs between increasing efficiency and reducing disparities, where the latter may relate to objectives such as the longer-term health of the marketplace or the organization's mission. We examine how participants' profile pictures shape this trade-off in the context of a peer-to-peer lending platform. We develop and apply an approach to estimate marketplace participants' preferences for different profile features, distinguishing between (i) "type" (e.g., gender, age) and (ii) "style" (e.g., smiling in the photo). Relative to type, style features are easier to change, and platforms may be more willing to encourage such changes. Our approach starts by using causal inference methods together with computer vision algorithms applied to observational data to identify type and style features of profiles that appear to affect demand for transactions. We further decompose type-based disparities into a component driven by demand for certain types and a component that arises because different types have different distributions of style features; we find that style differences often exacerbate type-based disparities. To improve internal validity, we then carry out two randomized survey experiments using generative models to create multiple versions of profile images that differ in one feature at a time. We then evaluate counterfactual platform policies based on the changeable profile features and identify approaches that can ameliorate the disparity-efficiency tension.
MetaX prepares a launch in Hong Kong to finance its battle against Nvidia | Cloud News
MetaX aims to accelerate its transition from a rising Chinese GPU manufacturer to an artificial intelligence computing platform. The company, based in
AI in Supply Chain Management Market Size, Trend & Growth | 2030
The AI in supply chain management market size was valued at USD 3.5 billion & is likely to grow at a CAGR of 30.3% during 2024-2030.
Google found liable for bad AI Overview results. Let's play Truth Or Consequences
Hush, children, what's that sound? Has the flood gates' key been found?
AI demands more engineering discipline. Not less
That question was answered decisively last November. Ever since Opus 4.5 came out, AI has been able to generate code that is approximately as good as that of the median software engineer, at least for common patterns, and much faster and more cheaply.
From User to Architect - by Adam Pryor - Purposeful AI
Most people experience AI as a tool that helps them do things faster. You write a prompt, you get an output, you edit it, you move on. The AI is a very fast assistant sitting beside you, and you are still the one doing the work.
Cognitive Debt: AI as Intellectual Leverage and the Dynamics of Systemic Fragility
arXiv:2606.15078v1 Announce Type: new Abstract: We develop a formal theory of cognitive debt: the stock of unverified reasoning obligations that accumulates when individuals use AI as a substitute rather than a complement for first-principles cognition. The model features two state variables per agent, cognitive capital and cognitive debt, and a multiplicative production technology in which cognitive capital functions as collateral that determines the return to AI adoption. We establish six propositions. Rational agents incur positive cognitive debt because the costs are deferred, partially external, and masked by short-run productivity gains. Tranquil periods lower subjective risk assessments, raise AI substitution intensity, and compound leverage, generating a cognitive Minsky moment in which subjective risk falls while true systemic fragility rises. Expected crisis losses are convex in aggregate leverage. Post-crisis, output-target pressure can produce a false-correction loop in which agents patch AI failures with more AI. The decentralised equilibrium over-adopts substitutive AI relative to the social optimum because of systemic risk, cognitive public goods, and arms-race externalities. In a two-type heterogeneous-agent economy, high-cognitive-capital agents adopt AI more intensively and may eventually erode their unaided cognitive capital below that of initially lower-skilled agents.
LearnOpt: Recovering the Latent Cognitive Structure of Standardized Examinations via Knowledge Graphs and Constrained Optimization
arXiv:2606.15349v1 Announce Type: new Abstract: Standardized examinations are typically treated as uniform syllabus coverage problems. We argue they are better understood as adversarial systems with stable latent cognitive structures diverging systematically from official syllabi. We introduce LearnOpt, which recovers this structure from historical question papers and generates personalized, time-bounded study plans. Applied to nine years of NEET questions (2016-2024, n=1,496), LearnOpt builds an exam knowledge graph from LLM-tagged questions, extracts a five-category latent skill distribution, and formulates study planning as a knapsack-variant optimization over prerequisite-aware subgraphs with Bayesian Knowledge Tracing. Central finding: NEET's latent skill distribution is stable within a syllabus regime (consecutive-year KL divergence 0.004-0.032 for 2016-2021, non-significant under permutation testing) but shifts significantly with NCERT's 2023 syllabus rationalization: pooling 2016-2021 (n=1,072) vs 2023-2024 (n=392) gives KL=0.040 (p=0.0005), with Elimination/Negation questions rising from ~20-29% to ~31-35%. Latent structure, while not permanently stationary, is piecewise stable, with shifts detectable and attributable to curricular events. Within either regime, subject predicts skill profile more strongly than year. An optimization evaluation, using one real and two synthetic mastery profiles, shows the skill-weighted objective produces a modest but real reordering of recommended topics over a mastery-conditioned frequency baseline. Applying the pipeline to JEE Advanced reveals a profile dominated by Multi-concept Integration (80.9% vs. 33.3% for NEET), with a JEE-vs-NEET divergence (KL=0.505) exceeding NEET's largest cross-subject divergence: exam tier shapes latent cognitive structure more than subject, which shapes it more than time within a regime. Code, knowledge graph, and annotated dataset are released publicly.
Venture Capital in the Age of Wars: Geopolitical Competition and the Transformation of Innovation
The emergence of an “Age of Wars” is fundamentally reshaping venture capital. Technologies associated with security, resilience, strategic autonomy, and military effectiveness are increasingly replacing consumer-oriented innovation as the primary destinations for investment.
Swedish AI patent platform Lightbringer raises €8.6 million to “take on Big Law” and replace existing patent firms
Lightbringer, a Malmö-based AI-powered LegalTech company transforming how startups and SMEs secure patents, has raised €8.6 million ($10 million) in Series A funding to fuel its US expansion and next phase of product development. The round was co-led by London-based 6 Degrees Capital and Amsterdam-based Newion. Thomas Olszewski, Partner at 6 Degrees Capital, and Dorus […]
Amsterdam’s Anterra Capital hits €86 million first close for Fund III as AI reshapes food and agriculture
Anterra Capital, an Amsterdam-based specialist venture firm investing in food and agriculture, has announced the first close of its Fund III at €86 million ($100 million), against a target of €172.1 million ($200 million). Anterra’s investor base spans institutional investors, food system operators and industry innovators across North America, Europe and APAC, including Rabobank, Novo […]
Exclusive: AI Startup ContraVault Raises $3.1 Mn To Expand In The US
ContraVault AI helps businesses across construction, energy, power, defence and aerospace streamline the tender bidding process
Indian AI startups pitch for funds from venture capitalists in US
Singapore: More Indian startups ... for the domestic and international markets, an AI startup said. “We are seeing an increasing number of Indian AI companies pitching for funds in San Francisco, where venture capitalists and equity investors have been closely monitoring the ...
Labor, Society & Culture
Global Mobility of Highly Skilled Talent Falls Nearly 12% as Competition for AI Expertise Intensifies
/PRNewswire/ -- The international movement of highly skilled professionals fell sharply in 2025, with cross-border relocations dropping from 3.7 million to 3.3...
HR must manage AI bots as well as humans, says Accenture executive
Matt Prebble says businesses will be forced to rethink leadership models
Marketing jobs are among the most exposed to AI. Adobe and LinkedIn are teaming up to ensure the industry is upskilled—not replaced
In a Fortune exclusive, Adobe and LinkedIn announce new coursework designed to teach marketing professionals how to properly use AI.
Supply chain roles requiring AI skills outpacing overall labor market - Truck News
Demand for supply chain roles requiring artificial intelligence skills has increased 387% in just three years, significantly outpacing overall labor
Neeti Sharma: As Companies Turn to AI, Will Future Leaders Emerge Without Middle Management?, ETHRWorld
Neeti Sharma: Explore the impact of AI-driven efficiency on middle management and its implications for future leadership development in organizations. Experts warn of potential leadership gaps as companies streamline management layers.
The 3 skills most likely to survive AI automation
Perplexity’s Dmitry Shevelenko discusses the company's strategy and identifies the three most durable human skills for the future of work.
The Perils of Agency: How Developers Perceive, Prioritize, and Address Risks in Agentic AI Products
arXiv:2606.15485v1 Announce Type: new Abstract: Agentic AI systems act autonomously, use tools, adapt to context, and operate in complex real-world environments. However, these same characteristics can create or exacerbate product risks. We studied how industry developers (n=35) perceive, prioritize, and address the risks in their agentic AI products. We found that developers' perceptions of risk were closely tied to the qualities that made the product agentic, such as autonomy, tool use, and usage in a real-world context. Developers prioritized product and business risks before considering downstream societal risks like job displacement and end-user privacy. This prioritization also impacted developers' ability and motivation to mitigate agentic risks. Finally, developers lacked mature controls for containing agentic risks, often relying on constraining the same characteristics that make agents useful: e.g., autonomy and goal complexity. These findings reveal a capability vs. risk control tension in agentic AI development: developers need to address risks that emerge from agentic capabilities, yet they currently have limited support for doing so without constraining agentic functionality.
Europe’s AI champion Mistral vulnerable to Russian disinformation, study finds
Open-source generative models are worse at removing false news than others, according to Estonian researchers
OSGuard: A Benchmark for Safety in Computer-Use Agents
arXiv:2606.15034v1 Announce Type: new Abstract: Computer-use agents are increasingly evaluated by whether they complete realistic desktop and web tasks. However, task success alone can miss failures in which an agent reaches the nominal goal through an unsafe shortcut. We introduce OSGuard, a dual-granularity benchmark suite for evaluating safety in computer-use agents under benign, unchanged user instructions. OSGuard contains an action-level benchmark for local guardrail decisions and a risk-augmented execution suite for end-to-end evaluation. The action-level benchmark consists of contextualized proposed actions labeled as allowed, unrelated, or unsafe, each judged relative to the original instruction and current interface state. The execution suite contains manually constructed OSWorld-derived task variants in which the original task remains achievable, but the environment is modified to introduce latent hazards such as destructive overwrites, etc. Each variant is paired with augmented evaluators that retain the original task-success criterion while adding explicit state-based safety invariants, allowing us to distinguish safe completions from unsafe completions that satisfy the nominal task objective. Our experimental results on OSGuard show that current multimodal guardrails can perform well on isolated action judgments, while risk-augmented execution exposes remaining gaps between local oversight and reliable end-to-end safety. This dual-granularity design enables more precise diagnosis of whether models can both recognize unsafe proposed actions and improve full-task safety when deployed as guardrails.
From Distorted Mirrors to Sovereign Reflections: Resisting the Grotesque Depiction of Our Digital Selves
arXiv:2606.15728v1 Announce Type: new Abstract: As Ambient Intelligence weaves computing into everyday life, human existence has become inextricably linked to ubiquitous digital infrastructures, triggering a crisis of personal data sovereignty. Driven by extractivism and data colonialism, platforms exploit human experience as a terra nullius, enclosing proprietary user models within isolated silos. We argue that legal frameworks like the EU Digital Rights and GDPR fail to counter this power asymmetry because they regulate inert raw data while leaving monopolies over algorithmic inferences unchallenged. Consequently, current architectures act as funhouse mirrors, creating a grotesque, distorted depiction of our digital selves. Through critical posthumanism and four contemporary scenarios (entertainment, e-commerce, fintech, and telematics), this paper illustrates how data siloing inflicts cognitive amputation, ontological violence, and self-censorship. We argue that securing digital rights requires moving beyond defensive regulation toward a decolonial infrastructure that guarantees the functional transferability of algorithmic identities, reclaiming technology for the emancipation of the posthuman self.
The skills young financiers will need to thrive in the age of AI
Mastering finance at a time of technological disruption will rely on human abilities
69% Education leaders flag curriculum-industry gap, 71% see AI shaping future of learning: ETEducation survey
ETEducation White Paper 2026: Exclusive ETEducation survey of 300+ education leaders across 15+ cities reveals employability, industry alignment and AI adoption as key priorities shaping India’s education sector towards 2035.
Gender Differences in AI Literacy Workshop Outcomes and Deepfake Engagement
arXiv:2606.14718v1 Announce Type: new Abstract: As Artificial Intelligence (AI) literacy initiatives expand in K-12 settings, understanding how gender shapes student baseline perceptions, tool-use, and responsiveness to interventions is essential for equitable curriculum design. This study examines gender differences in AI literacy, safety awareness, and STEM career aspirations among Australian secondary students (Years 7, 8, and 10; N(pre) = 199, n(post) = 136) from two co-educational government schools who participated in a one-day AI literacy workshop. Using statistical regression methods controlling for year level and school, we found that pre-workshop, male students reported significantly higher STEM career interest across all three domains (AI, computer science, and engineering), while female students were significantly more likely to use AI for schoolwork and to seek advice from AI tools. Gender-differentiated patterns also emerged in deepfake behaviours: males were significantly more likely to have created or shared deepfake content. Both genders improved in AI knowledge post-intervention, yet females showed a richer profile of gains: wider conceptual understanding, greater confidence, and meaningful increases in AI and CS career interest that partially narrowed the gender STEM gap. These findings highlight the need for gender-responsive AI curricula, particularly deepfake safety education for male students, and demonstrate that even single-day workshops can narrow gender gaps in STEM aspirations and AI confidence.
Agencies are doubling down on AI upskilling, but they may be solving the wrong problem | Federal News Network
Employees need to feel safe when they are practicing new things or when they are learning new things," said Priyanka Dave.
Technology & Infrastructure
Trust Between AI Agents: Measuring Formation, Breakage, and Recovery, with Implications for Governing Multi-Agent Systems
arXiv:2606.14923v1 Announce Type: new Abstract: As language-model agents increasingly work in teams, each agent must decide how much to trust its teammates. Yet we lack a standard way to measure trust between AI agents. We propose a behavioral measure based on costly verification. In a cooperative survival game, checking a teammate's work consumes resources, while trusting a wrong answer can be fatal. Relative to a memoryless version of the same model, reduced verification provides an observable measure of trust. Using this framework, we study trust formation, breakage, and recovery across six frontier model snapshots. When paired with a consistently reliable teammate, four snapshots (Claude Opus 4.6, Claude Sonnet 4.6, GPT-5.1, and Gemini 3.1 Pro) reduce verification by roughly 60-85%, whereas two smaller snapshots show little or no such adjustment. Failures reverse this discount, but models differ in how they respond. Some concentrate renewed scrutiny on the culprit, while others become more cautious toward the entire team. Recovery is slower than formation, and clustered failures sustain suspicion far longer than the same number of failures spread apart. These differences have practical consequences. Models that form trust verify less, decide more quickly, and achieve higher payoffs in our environment. By contrast, persistent over-verification is associated with indecision rather than safety. Our results show that trust dispositions can be measured before deployment and suggest that calibration, rather than maximal suspicion, should be the central concern in the governance of multi-agent AI systems.
Vibe coding can build your pipeline. It can't explain it six months later
AI coding agents are rapidly accelerating data engineering by generating transformations, pipelines, orchestration workflows, validation tests, and infrastructure configurations from prompts. However, enterprise data platforms have long operated across fragmented systems owned by different teams and built on different technologies. As these systems evolve independently, organizations increasingly struggle with inconsistent business logic, duplicated implementations, difficult downstream impact analysis, and hidden dependencies across the platform. The rise of vibe coding can further amplify these problems as more operational context, architectural decisions, and business knowledge become scattered across prompts, conversations, generated code, and disconnected workflows rather than becoming part of the system itself. Spec-driven development (SDD) is emerging as one approach to address this challenge. In SDD, prompts, business rules, validation logic, orchestration behavior, and implementation workflows are converted into executable and versioned specifications that become part of the system itself. These specifications act as persistent operational memory for both humans and AI agents, allowing systems to evolve more consistently across releases, teams, and AI-assisted workflows. Because enterprise data engineering already relies heavily on reusable patterns, metadata-driven pipelines, and standardized operational workflows, it is especially well-suited for SDD. By combining AI-assisted generation with deterministic and reusable system contracts, SDD may provide a new operational layer for reducing fragmentation and improving long-term coordination across increasingly AI-generated data platforms. Vibe coding alone lacks persistent system memory Vibe coding works remarkably well for generating isolated implementations quickly. But prompts are inherently temporary. They capture an engineer’s assumptions, business context, implementation logic, and system knowledge only for that specific conversation and moment in time. In practice, making AI-generated systems work often requires far more than a simple prompt. Engineers continuously provide background information, architectural decisions, business rules, schema assumptions, downstream dependencies, operational constraints, debugging history, and implementation guidance throughout the development process. These contexts become the real operational knowledge behind AI-assisted development. However, in most vibe coding workflows, this information remains scattered across prompts, conversations, Jira tickets, documentation, chat history, generated code, and disconnected workflows rather than becoming part of the system itself. This creates a major problem for enterprise data engineering because modern data platforms are naturally fragmented across many interconnected systems, including ingestion pipelines, warehouses, orchestration frameworks, semantic layers, APIs, dashboards, and machine learning (ML) systems. As more logic and context become embedded inside prompts and generated implementations, organizations gradually lose visibility into: architectural intent downstream dependencies validation assumptions operational behavior business context behind implementations Over time, the system itself no longer contains the full reasoning behind how it was built. Critical business context, architectural assumptions, and operational knowledge still largely exist inside human judgement and scattered conversations rather than inside the platform itself. Vibe coding makes implementation significantly faster, but from a system perspective, overall engineering efficiency does not improve proportionally because much of the development lifecycle still depends on human validation, domain knowledge, coordination, and decision-making. More importantly, prompts are not naturally iterable engineering artifacts. Enterprise systems continuously evolve across releases, schema changes, business logic updates, and downstream dependencies. Teams repeatedly revisit and refine systems over time, but prompts are optimized for fast local generation rather than system long-term evolution. They are difficult to: version consistently validate systematically reuse across teams coordinate through CI/CD workflows evolve incrementally over time Even the same prompt may not reliably generate the same implementation with different context in the future. This is where SDD begins to move to the center of AI-assisted data engineering. Instead of leaving operational knowledge scattered across prompts and conversations, SDD integrates business context, validation logic, transformation behavior, orchestration requirements, and implementation workflows directly into executable specifications that become part of the system itself. The system now has persistent memory about how it was designed, why certain decisions were made, and how different components are connected across the platform. This allows teams and AI agents to iterate systems more reliably over time while reducing fragmentation across increasingly distributed data environments. Spec-driven development turns prompts into system memory In SDD, systems are built around executable specifications rather than loosely coordinated prompts and implementations alone. Instead of treating specifications as passive documentation written after development, SDD treats them as operational contracts that directly drive code generation, validation, testing, orchestration, and deployment workflows. In many ways, SDD extends ideas from Infrastructure-as-Code and GitOps into AI-assisted engineering. Specifications combine declarative system definitions with executable implementation workflows. The declarative layer provides system context, schemas, dependencies, constraints, and operational requirements, while workflow-oriented instructions guide AI agents on how to implement and evolve the system consistently. Once these contexts, rules, and implementation patterns are converted into persistent and versioned contracts stored in repositories and integrated into CI/CD workflows, the system becomes significantly more iterable and governable over time. These specifications effectively become long-term system memory for both humans and AI agents, allowing systems to evolve consistently across releases, teams, and increasingly AI-assisted development workflows. In practice, the structure of specifications largely depends on the type of systems and workflows being implemented. However, spec-driven systems often begin with a foundational “constitution” that defines project-wide principles and constraints that should remain consistent across the platform, such as technology standards, naming conventions, architectural rules, governance policies, and core system requirements. On top of this foundation, multiple layers of specifications serve different operational purposes across the development lifecycle: schema specifications define structural compatibility transformation specifications define business logic validation specifications define quality rules orchestration specifications define execution behavior semantic specifications define shared business definitions AI workflow specifications define reusable implementation instructions for coding agents A simplified specification might look like this: pipeline_spec: source: system: mysql table: order transformation: logic: - load_strategy: scd2 target: platform: snowflake table: dim_order validation: primary_key: order_id Additional workflow files can then provide reusable implementation instructions for coding agents: Generate Python ingestion code for Salesforce customer data. Generate DBT models implementing Type 2 SCD logic. Generate Airflow workflows for hourly execution. Generate validation tests for downstream compatibility. These specification documents are often maintained as markdown-based operational artifacts generated and refined through AI-assisted workflows. Engineers can iteratively update the specifications, provide additional business context, and collaborate with coding agents to improve implementation logic, workflows, and prompt instructions over time. Compared to traditional documentation processes, AI-assisted specification generation is significantly faster and more adaptive. The important shift is not simply better documentation. Specifications become reusable operational context that allows systems to evolve consistently across releases, teams, and AI-assisted workflows. Architectural intent, business assumptions, and implementation logic no longer disappear into temporary prompts and disconnected implementations, but instead become persistent system knowledge integrated directly into the development lifecycle. Why spec-driven development specifically fits data engineering SDD can theoretically be applied across many areas of software engineering, but data engineering is especially well-suited for this model because of the nature of modern data platforms. Enterprise data systems naturally span many interconnected technologies and layers, including transactional systems, ingestion frameworks, streaming platforms, warehouses, orchestration systems, semantic layers, APIs, dashboards, and ML pipelines. Data engineers regularly work across long technology stacks and distributed systems where a single upstream change can impact many downstream consumers. Enterprise data platforms also support many different teams and applications across fragmented environments. As systems evolve independently, understanding the full downstream impact of an upstream schema or business logic change becomes increasingly difficult. A seemingly small modification can silently break downstream pipelines, dashboards, APIs, semantic models, or machine learning workflows across the platform. SDD can address this fragmentation by introducing shared and versioned operational contracts across systems. Because schemas, dependencies, validation rules, transformation logic, and orchestration behavior are explicitly defined within specifications, teams and AI agents gain much better visibility into how systems are connected and how changes propagate across the platform. Additionally, the goal of data engineering is not simply delivering pipelines quickly. Teams must also optimize for system stability, scalability, consistency, maintainability, operational reliability, and infrastructure cost. This requires significant system and solution design work from engineers. Teams must define tech stack, create schemas, transformation patterns, orchestration behavior, validation rules, storage strategies, and downstream compatibility requirements carefully across the platform. However, once these architectural and operational patterns are established, much of the implementation work becomes highly repetitive and standardized. For example, after defining a reusable ingestion and transformation pattern for Salesforce customer data, onboarding a new table may only require adding another table definition into the specification, while the remaining implementation can be generated automatically through existing specifications and workflows that follow the same operational pattern: source: system: salesforce tables: - customer - order - product From this specification alone, coding agents could generate new data pipelines following the same governed implementation pattern across the platform. This combination of human-driven architectural design and highly repeatable implementation workflows makes data engineering particularly suitable for SDD. In many ways, data engineering has always been moving toward higher levels of automation, from ETL frameworks and metadata-driven pipelines to IaC and declarative orchestration systems. SDD represents another step in that evolution by combining prompt-based AI generation with deterministic and versioned operational contracts. Instead of relying entirely on temporary conversational prompts or rigid template systems, SDD introduces a middle layer where reusable specifications provide structure, coordination, validation, and persistent system memory for AI-assisted development. How SDD changes AI-assisted data engineering SDD introduces a much higher level of automation into enterprise data engineering while also helping reduce the fragmentation problems that modern data platforms increasingly face. Because schemas, business rules, transformation behavior, orchestration requirements, validation logic, and downstream dependencies are explicitly defined inside reusable specifications, coding agents can generate and evolve large portions of the implementation consistently across the platform. Instead of repeatedly rebuilding pipelines and workflows from temporary prompts and disconnected context, teams can iterate systems through shared operational contracts and reusable implementation patterns. This significantly improves consistency, traceability, and coordination across distributed environments. Schema evolution becomes easier to manage, downstream impact becomes more visible, and systems can evolve incrementally instead of through disconnected generations of implementations. At the same time, human engineers still remain essential in the development lifecycle. While AI agents can automate large portions of implementation work, human judgement is still critical for defining business logic, designing architectures, managing tradeoffs, validating correctness, and coordinating system evolution across organizations. As more implementation work becomes AI-generated, the role of data engineering also begins shifting. Engineers spend less time writing repetitive pipelines and orchestration logic, and more time defining specifications, designing reusable operational patterns, managing validation rules, and coordinating business context across systems. This may also gradually reduce some of the traditional boundaries between different data engineering teams. Because implementation becomes increasingly standardized and AI-assisted through shared specifications, organizations may rely less on highly siloed platform-specific implementation teams and more on shared operational contracts and reusable system patterns. Ultimately, SDD shifts data engineering toward a more specification-oriented and system-oriented model where humans focus on intent, architecture, and business coordination, while AI agents increasingly handle implementation, testing, and operational generation at scale. Shuhua Xu is a lead data engineer.
PrologMCP: A Standardized Prolog Tool Interface for LLM Agents
arXiv:2606.14935v1 Announce Type: new Abstract: Frontier reasoning-tuned language models still fail on deductive tasks at depth, and the cost of improved performance through extended internal reasoning scales poorly. Symbolic delegation offers a complementary route: a language model translates the problem, while a solver performs the inference. However, current autoformalization pipelines for logic programming are typically bespoke integrations tied to particular tasks or agents. We introduce PrologMCP, a task-agnostic, open-source server that exposes Prolog as a stateful tool through the Model Context Protocol (MCP). Its compact tool interface, structured error reporting, and per-session isolation make the translate-run-inspect-repair loop a reusable primitive for MCP-capable agents. We evaluate a formalizer agent enhanced with PrologMCP against standard and reasoning LLMs (Claude Sonnet 4.6, GPT-4.1, and o4-mini) on two subsets of PARARULE-Plus: a general-purpose sample and a more challenging one targeting a specific failure mode of natural-language reasoning. On the general sample, the formalizer matches or exceeds reasoning LLMs (accuracy 1.00 vs.\ 1.00 / 0.998), with the largest gains over standard models (0.762 for GPT-4.1). On the challenging subset, the formalizer remains near-perfect (1.00 / 0.99) while reasoning LLMs drop to 0.95 / 0.94. These results suggest that delegating inference to Prolog via MCP is a robust and inspectable alternative to extended natural-language reasoning.
Agentic AI systems are doing more and more work. Now humans need to figure out how to verify it all
At Fortune Brainstorm Tech, industry executives discussed the challenges and techniques for bringing accountability into AI.
Shachi: A Modular, Controllable Framework for LLM-Based Agent-Based Modeling of Emergent Collective Behavior
arXiv:2509.21862v3 Announce Type: replace-cross Abstract: How collective behaviors emerge from the interactions of individual LLM-driven agents is a central question in artificial life, yet controlled study of these emergent dynamics has been hindered by the lack of a principled simulation framework for systematic experimentation. To address this, we introduce Shachi, a principled methodology and modular framework that decomposes an agent's cognition into core components: Configuration for intrinsic identity, Memory for contextual continuity, and Tools for extended capabilities, all orchestrated by an LLM reasoning engine. This decomposition treats each cognitive component as an independently controllable variable, enabling perturbation studies that trace how micro-level cognitive traits propagate into population-level dynamics. We investigate behavioral patterns across a 10-task benchmark spanning three levels of collective complexity. Shachi enables memory transfer across environment transitions, producing history-dependent behavioral shifts, and allows agents to simultaneously inhabit multiple environments, revealing cross-environment interference invisible in single-environment studies. Furthermore, in a real-world U.S. tariff shock case study, locally interacting agents with individually controlled cognitive components produce macro-level market dynamics directionally consistent with observed real-world outcomes. Our work provides a rigorous, open-source simulation framework for LLM-based ABM, aimed at fostering cumulative scientific inquiry into the emergent collective behaviors of interacting artificial agents.
Risk-Aware LLM Agents for Geospatial Data Retrieval: Design and Preliminary Adversarial Evaluation
arXiv:2606.15077v1 Announce Type: new Abstract: We present an LLM-driven framework for retrieving remote sensing data from cloud-based geospatial catalogues using natural language queries. The system converts user intent into structured API calls, enabling efficient access to satellite imagery and environmental datasets. The architecture integrates three agents: Guardrail for safety and policy enforcement, General-QA for intent interpretation, and Recommender-Analyst for schema-aware API call generation. This coordinated design ensures reliable, semantically aligned interaction with external data services. The modular framework is portable across platforms through API schema substitution and supports applications in environmental monitoring, disaster response, and climate analysis. It establishes a scalable interface between user intent and geospatial infrastructure, enabling streamlined and automated Earth observation workflows. Preliminary experiments under adversarial multi-turn settings show that prompt-level safety instructions improve robustness, although rare high-impact failures persist in API manipulation scenarios and highlight the need for adaptive, system-level defenses that balance safety, usability, and cost efficiency, which motivates the use of our intercept-level Guardrail agent.
China's Nexchip breaks into foundry top eight after AI demand lifts market to record numbers
AI, high-performance computing, and early pull-ins from TV, PC, and notebook supply chains pushed the global foundry market to a record high in the first quarter of 2026. China's Nexchip Semiconductor delivered the key ranking shift, overtaking Taiwan's Vanguard International Semiconductor ...
Two mayors, one $10 billion AI data center, and a growing divide in small-town Texas
What happens when a megaproject lands next door—but your town has no authority to stop or shape it?
CONCORD: Asynchronous Sparse Aggregation for Device-Cloud RAG under Document Isolation
arXiv:2606.15179v1 Announce Type: new Abstract: Retrieval-augmented generation (RAG) has emerged as a pivotal technique for improving language models by incorporating external knowledge at inference time. As device-cloud collaborative inference makes it feasible to deploy small language models on edge devices, a new setting arises in which private documents remain on the device and public knowledge resides in the cloud. Privacy and policy constraints often forbid raw document exchange, creating a document-isolated dual-end RAG setting. However, existing methods rely on frequent remote synchronization and dense evidence transfer, limiting throughput under realistic latency and bandwidth conditions. To address this issue, we propose CONCORD, an asynchronous sparse aggregation framework for dual-end RAG under document isolation. CONCORD treats the cloud as an asynchronously arriving evidence source rather than a continuously synchronized co-generator. Specifically, we introduce waiting debt control to decide whether each decoding step should continue waiting for remote participation based on the observed return of waiting. We also design a certificate-guided minimal supplementation mechanism that requests only the remote evidence needed to determine the current greedy decision. Steps that consult the cloud preserve the same greedy token as dense dual-end aggregation, while the remaining steps commit locally without remote evidence. Experiments on Natural Questions and WikiText-2 show that CONCORD improves end-to-end throughput over baselines by $1.66\times$ and $2.15\times$, respectively, while reducing per-token communication by over two orders of magnitude and maintaining comparable answer quality and perplexity.
Google Commits US$50M to Tackle Construction Skills Gap | Construction Digital
As data centre demand soars, Google has committed US$50m to skilled trades training, days after Meta made a US$115m pledge to solve the same crisis
Global data center energy consumption rivals entire countries. Here’s what to know about what they mean for your electricity and climate as they boom in Tennessee.
Data centers have been around for decades, but according to experts, many of the facilities go beyond what has been traditionally seen in the U.S. — and can come with a steep cost.
AWS rolls the dice for faster, more efficient networking
Honey, I flattened the datacenter network.
Towards Verifiable Agentic Data Science: Solving Irregular TSQA Via Tool-Grounded Reasoning
arXiv:2606.15107v1 Announce Type: new Abstract: Time series data in real-world deployments is overwhelmingly irregular. Observations are asynchronous, missing values are informative rather than random, and sampling frequencies vary across sensors and operational windows. However, existing Time Series Question Answering (TSQA) benchmarks mostly assume regularly sampled inputs, leaving a fundamental gap in understanding how large language models (LLMs) and AI agents perform under irregular conditions. To bridge this gap, we introduce IRTS-ToolBench, a benchmark of 1,700 questions spanning 10 task types across 13 domains. IRTS-ToolBench is designed to be used independently by any researcher working on LLM-based irregular time series analysis, providing standardized inputs and a reproducible evaluation protocol. Code can be found in https://github.com/SanhornC/IRTS-ToolBench.
Relational Structural Causal Models
arXiv:2606.14892v1 Announce Type: new Abstract: An artificial intelligence must have a model of its environment that is causal, supporting reasoning about interventions and counterfactuals, and also combinatorial, supporting generalization to unseen combinations of objects. In this work, we formally study when and how such a model can be learned. We develop relational structural causal models, extending structural causal models (Pearl 2009) to settings where objects and their relations vary. First, we show how answers to not only causal but also observational queries about unseen combinations of objects can not be identified without further assumptions. To enable such identification--including in the presence of unobserved confounding--we define relational causal graphs and derive symbolic identification criteria. Finally, we propose relational neural causal models, a provably correct approach that outperforms non-relational baselines on simulated traffic scenes with varying cars, signals, and pedestrians.
AI Engram: In Search of Memory Traces in Artificial Intelligence
arXiv:2606.14997v1 Announce Type: new Abstract: Memory formation is fundamental to intelligence, yet whether deep neural networks preserve identifiable memory traces analogous to biological memory units remains an open question. This work introduces a geometric framework to identify such "AI engrams" by formalizing the neuroscientific criteria of specificity, reactivation, sufficiency, and necessity into a constrained inverse problem. We derive a closed-form estimator that isolates individual memory traces from globally entangled parameters, and show that this biologically-derived solution corresponds to a natural gradient update on the parameter manifold. AI engrams enable surgical manipulation of learned knowledge: any subset of memories can be composed or erased through linear arithmetic, without iterative optimization. Experiments ranging from simple MLPs to LLMs demonstrate the causal validity and substantial scalability of AI engrams. Together, these results bridge theories of biological memory and artificial representation learning and offer geometric insight into how deep networks simultaneously support functional specificity within distributed storage.
Vision LLMs are PDF Parsers Too: Reading Charts and Diagrams for RAG
How vision-capable large language models can be used to extract data from complex PDF documents.
Adoption, Deployment & Impact
A Definition of Good Explanations and the Challenges Explaining LLM Outputs
arXiv:2606.14838v1 Announce Type: new Abstract: How to define a good explanation is a long-standing philosophical debate which has found recent renewed interest in the context of AI outputs. Explainability is crucial for AI adoption in many contexts, but in order to produce good explanations of AI systems, we must first have an understanding of what good explanations are. In this paper we propose a definition inspired by the notion of counterfactual explanations, however we argue that one must also take into account the interlocutor's prior beliefs in each fact that could be offered in an explanation. We explore the ramifications of this definition for AI explainability and, in particular, why LLM outputs are difficult to produce good explanations for.
Brokerage AI adoption rises, productivity gains remain uneven
Brokerages report 97% AI adoption, while agents use AI mainly for marketing, with gains concentrated among power users, per RPR.
What Building 375 AI Agents in Five Days Revealed About Where Enterprise AI Adoption Breaks Down
Opal University reinforced a reality many enterprises are still underestimating: AI adoption does not scale simply because companies buy more tools.
Mind the Marketing Gap: Most CMOs Say AI Is Transforming Marketing, But Few Are Using It to Transform Their Own Function
/PRNewswire/ -- The vast majority of chief marketing officers (CMOs) feel that AI is driving an end-to-end transformation of their function. But only 8% are...
Survey: Nearly All European Organisations Feel Pressure to Scale AI for Customer Experience, Yet Only 38% Have a Clear Approach to Governance
Survey of 200 senior leaders across Western and Central Europe reveals widening gap between AI adoption and governance exposing risks in compliance, multilingual CX and customer trustNOTTINGHAM, England--(BUSINESS WIRE)--New research from CallMiner, t...
How Enterprises Are Re-Architecting Data For The Agentic AI Era
Inc42 and Skyflow hosted a closed-door roundtable on scaling agentic AI safely, bringing together tech architects and engineering heads
Most Entrepreneurs Are Adding AI in the Wrong Order. Here's the Framework That Actually Works | White Beard Strategies
Here is the framework most AI consultants won't tell you about, because it means selling you fewer tools: before you add a single AI platform, you need to
Optimising Temporary Accommodation Placement Across London with AI-Powered SaaS in E-Governance Systems
arXiv:2606.16652v1 Announce Type: new Abstract: Temporary accommodation has become a major fiscal and administrative pressure for English local authorities, particularly in London, where demand and costs have risen sharply. This paper documents the creation and use of DOMUS, a cloud-based, AI-enabled decision-support system built from scratch at the University of East London and customised for the needs of London Borough of Newham to support statutory Temporary accommodation placement. DOMUS integrates household case records, policy-constrained affordability and suitability rules, and live private-rental listings within a single governance-aligned workflow. The system combines transparent, rule-based filtering with large language model-assisted search to standardise the application of bedroom need, affordability thresholds, geographic preferences, and accessibility requirements, while preserving officer discretion and audibility. Household and property attributes are encoded into policy-consistent representations prior to AI-assisted ranking and explanation. A pilot deployment in Newham's secure environment evaluated operational performance relative to manual workflows. Results indicate substantial reductions in search time, improved adherence to key placement constraints, and high staff satisfaction, while maintaining statutory compliance and role-based accountability. Beyond TA, the paper frames DOMUS as replicable digital public infrastructure: a modular, cloud-native Software-as-a-Service architecture that can be deployed across other UK boroughs and adapted to other public administration tasks characterised by scarcity, rule-bound eligibility, and high stakes. The findings demonstrate the feasibility of scalable, ethically governed AI deployment in local government and contribute to debates on AI-enabled public value creation in e-governance.
Fusion is not one-size-fits-all: Cross-Modal Representation Alignment for Time-to-Event Modeling
arXiv:2606.15038v1 Announce Type: new Abstract: Accurate time-to-event (TTE) prediction from multimodal clinical data remains challenging due to modality imbalance and distribution shift. We introduce a foundation model-driven framework for cross-modal representation alignment between CT imaging and longitudinal EHR data, designed to generalize across tasks and institutions. CT and EHR modalities are encoded independently using domain-specific foundation models and aligned in a shared latent space through four principled fusion strategies: late fusion, contrastive alignment, cross-attention, and co-attention. We evaluate two clinically distinct TTE tasks: pulmonary embolism (PE) mortality and cardiovascular disease (CVD) outcomes, on large-scale multi-institutional cohorts (PE: N=3,099 train; 1,098 internal; 435 external; CVD: N=2,951 train; 837 internal; 682 external). Fusion consistently improves concordance index by 1.5-5.4% over unimodal baselines when modalities contribute comparably. Overall, contrastive multimodal fusion, particularly with CLMBR representations, provided the most consistent and statistically robust improvements, especially for PE mortality prediction. For MACE, cross-attention (one-hot) achieved the highest internal performance and image-guided co-attention achieved the best external performance. We therefore introduce a generalizable foundation model-based cross-modal alignment framework and provide the first systematic analysis of fusion behavior under modality imbalance in TTE prediction. Our results establish task-aware multimodal alignment as a necessary design principle for robust generalization and scalable clinical deployment.
Semantics-Enhanced Retrieval-Augmented Time Series Forecasting
arXiv:2606.14941v1 Announce Type: new Abstract: Time series forecasting models often benefit from historical patterns. Inspired by Retrieval-Augmented Generation (RAG), recent research explored retrieving relevant historical time series segments to enhance forecasting. However, relying solely on time series similarity is often insufficient for retrieval under non-stationarity. To address this, we propose a multimodal approach: a \textbf{S}emantics-\textbf{E}nhanced \textbf{R}etrieval-\textbf{A}ugmented Time Series \textbf{F}orecasting framework, SERAF. Unlike mainstream approaches that depend only on time series similarity, SERAF conducts dual retrieval over the time series and their self-generated textual descriptions. It retrieves two complementary sets of historical patterns and corresponding futures, which are selectively and jointly used to guide future predictions. Experiments across seven real-world datasets demonstrate the effectiveness of SERAF in bridging numerical and semantic views of time series compared with state-of-the-art baselines.
A Nationwide Benchmark for Wildfire Initial Attack Failure Prediction with Public Environmental Data
arXiv:2606.15529v1 Announce Type: new Abstract: Initial attack (IA) is the first wildfire suppression phase, when agencies must quickly decide which fires may escape early control. Existing IA failure prediction studies often use non-public response records or regional settings, so it remains unclear how well public data available at fire discovery time can support IA failure prediction at national scale. We present WILDFIREIA, the first U.S. national-scale benchmark for IA failure prediction from environmental and contextual data available at fire discovery time. WILDFIREIA aligns 38,128 naturally caused FPA-FOD wildfire events with FIRMS/VIIRS thermal detections, gridMET weather and fire-danger variables, LANDFIRE vegetation, fuel, and topography, OpenStreetMap access features, and WorldPop population density. To prevent data leakage, the benchmark fixes the event unit, size-based label rule, chronological split, metrics, and forbidden-feature list, and excludes final fire size, containment timestamps, and post-discovery satellite detections from model inputs. We evaluate 16 representative models across tabular, temporal, spatial, and spatiotemporal families under the same protocol. Results show that public discovery-time data provides useful but incomplete signal for IA failure prediction: XGBoost achieves the best AUPRC of 53.3%; FIRMS/VIIRS is the least redundant source; and fuel is the strongest static predictor when dynamic observations are unavailable. We release preprocessing outputs and model-ready caches to support reproducible research on early wildfire risk assessment: https://github.com/LabRAI/WildfireIA#.
Improving Capstone Team Outcomes through Dynamic Skill Matching and Preference Alignment
arXiv:2606.15572v1 Announce Type: new Abstract: Team-based projects are a cornerstone of engineering and computing courses, but unstructured team formation often leads to poor project outcomes due to misaligned student interests and inadequate skill coverage. This paper introduces a novel, three-stage methodology for creating effective student teams by integrating student preferences with project skill requirements. In the first stage, students complete a survey to report their project interests and self-assessed skills. Next, a Large Language Model (LLM) analyzes project descriptions to extract the necessary skills for each project's success. Finally, a dynamic assignment algorithm matches students to projects, simultaneously maximizing skill coverage and preference alignment. The algorithm iteratively prioritizes projects with unfulfilled skill needs to optimize team balance. Preliminary evaluations show our approach produces teams with higher skill coverage and better preference satisfaction compared to random or manual assignment approaches. Our approach also overcomes limitations of widely-used tools like CATME Team-Maker, which do not explicitly account for project skill fulfillment. Our findings point toward an effective and customizable strategy for improving student motivation and learning outcomes in project-based courses.
UK evaluates frontier AI for operational cybersecurity applications | Digital Watch Observatory
A new UK pilot demonstrated how AI can support cyber teams in finding critical weaknesses.
Irish fleet safety tech company CameraMatics raises €49m
The Dublin-headquartered company uses AI and analytics to improve road safety for commercial fleets. Read more: Irish fleet safety tech company CameraMatics raises €49m
Geopolitics, Policy & Governance
Towards a Theory of Modular Natives: Explaining Superscaling, China's Greatest Innovation Yet
arXiv:2606.15757v1 Announce Type: cross Abstract: First, we present a new theory of "modular natives." A modular native is a basic building block that is born modular, e.g., a solar cell. The theory predicts that using modular natives in building things reduces complexity and improves predictability, resulting in better outcomes and faster scale-up. Second, we test the theory on the largest dataset of its kind. We find, at a high level of statistical significance, that modular natives operate under a fundamentally different risk regime than other project types, with finite and predictable risk, in contrast to non-natives that have infinite and unpredictable risk. The findings help explain why modularity is key to successful building while bespokeness often leads to failure. Third, we relate our findings to economic and geopolitical development, arguing that China understands modular natives and scale-up better than any other geography and that this is key to China's swiftly growing dominance in renewables, batteries, EVs, robots, etc. We argue that China's mastery of modularity and scale-up is a major innovation in its own right, among the greatest and most impactful in human history, falsifying the common notion that China cannot innovate. Business and government outside China ignore these findings at their peril. Finally, we spell out policy and practice implications and identify areas for further research.
US embargo reignites global sovereign AI push after Anthropic models shut down
The abrupt suspension of Anthropic's Fable 5 and Mythos 5 models is reigniting sovereign AI discussions around the globe, delivering a stark warning that reliance on foreign tech providers can undermine national autonomy.
India needs a complete sovereign AI stack, not just models: Bessemer | Company News - Business Standard
Bessemer Venture Partners says India's AI ambitions must extend from compute and data infrastructure to agents, services and robotics
AI Nationalism Raises Dependence Risks Across Europe | Let's Data Science
Tyler Cowen and Alexander Tabarrok published a Marginal Revolution post on June 16, 2026 titled "AI nationalism, Europe included," warning that regional AI champions could create political leverage and cross-border dependence. The post uses a hypothetical about France's `Mistral` to illustrate ...
Are Government Stakes the Key to AI Sovereignty? by Angela Huyue Zhang - Project Syndicate
Angela Huyue Zhang identifies the competing US and Chinese strategies for AI sovereignty.
Sovereign AI strategies are converging on bottleneck blueprints (Analyst Angle)
As nations race to build sovereign AI, successful strategies are converging on a common formula: control critical bottlenecks. Vish Nandlall with the analysis.
Feds snooze as US datacenter law set to lapse with no replacement in site
Federal Data Center Enhancement Act (FDCEA) of 2023 covers standards including security and sustainability
Commons-Governed Artificial Intelligence: A Taxonomy of Collective Governance
arXiv:2606.15466v1 Announce Type: new Abstract: The governance of artificial intelligence is overwhelmingly theorized through two institutional frames. In the market frame, the data, models, and compute that constitute the AI stack are private goods exchanged under property and contract; in the state frame, a regulator imposes rules from above. A third possibility, the collective and self-organized stewardship of AI-relevant resources by the communities that produce and depend on them, remains comparatively under-theorized, even as it proliferates in practice through data trusts and cooperatives, federated learning consortia, public compute initiatives, open-weight model collaborations, and community data sovereignty regimes. This article argues that these arrangements form a coherent institutional family, which we call commons-governed artificial intelligence, and that the analytic vocabulary developed by Elinor Ostrom and her successors for common-pool and knowledge commons is the right backbone for classifying them. We contribute a two-dimensional taxonomy whose first axis is the resource layer of the AI stack held in common, distinguishing data, compute, models, knowledge and evaluation, and energy, and whose second axis is the governance function performed, derived from Ostrom design principles. We populate the taxonomy by examining the published evidence layer by layer, locate ten recurrent institutional archetypes within it, synthesize their positions through a maturity matrix and a comparative reading against the eight design principles, and treat the energy and sustainability of computation as a first-class commons-governance problem rather than an externality. We close with the tensions that constrain the project, openwashing, the compute bottleneck, free-riding, and the tension between scale and sustainability, and with a research agenda for a polycentric AI commons.
EU governments remain divided over plans to simplify digital rules
EU member states are split over proposals to simplify digital rules, with some countries arguing the latest draft does not sufficiently reduce regulatory burdens.
Anthropic's Fable 5 scramble
Anthropic's powerful Fable 5 AI model was pulled after an urgent report from Amazon triggered a White House intervention over national security concerns.
The EU has regulated itself out of the AI race but the UK is still in the game
Thanks to Brexit the UK isn't bound by Europe's instinctive 'safety first' approach to AI regulation - but there is still work we must do
AI Regulation Forum (2 days)
The AI Regulation Forum is a two-day event scheduled to take place in Brussels.
Chinese regulator launches reporting portal for AI-linked violations
The Cyberspace Administration of China has opened a new channel for reporting AI misconduct, including deepfakes and inadequate content labeling.
Denmark: Licensed Operators are Required to Assess the Risk of New Technology Prior to Launch
In Denmark, licensed operators are required to assess the risk of new technology prior to launch.
Get the full executive brief
Receive curated insights with practical implications for strategy, operations, and governance.