AI News Archive: May 25, 2026 — Part 11
Sourced from 500+ daily AI sources, scored by relevance.
- The pope takes on AI
Monday morning, the Roman Catholic Church made its biggest foray yet into the discourse on artificial intelligence and the role it should play in human life as the technology develops. In the first encyclical of his papacy, titled Magnifica humanitas (Latin for “magnificent humanity”), Pope Leo XIV argued that AI is not intrinsically immoral, but […]
- Pope Leo calls for being ‘profoundly human’ in the age of AI
Pope Leo XIV warned of the risks of AI and unconstrained technological power in his first major papal document released on Monday. Magnifica Humanitas is the pope's manifesto on "safeguarding the human person in the time of artificial intelligence," in which he discusses the dangers of AI-powered warfare, the effects of AI on labor, and […]
- Pope calls for robust regulation of AI in eagerly awaited first encyclical
Pope calls for robust regulation of AI in eagerly awaited first encyclical The Mercury News
- Pope calls for robust regulation of AI in manifesto that ponders the future of humanity
Pope calls for robust regulation of AI in manifesto that ponders the future of humanity Austin American-Statesman
- Pope calls for robust regulation of AI in manifesto that ponders the future of humanity
Pope calls for robust regulation of AI in manifesto that ponders the future of humanity Boston Herald
- Pope calls for robust regulation of AI in manifesto that ponders the future of humanity
Pope calls for robust regulation of AI in manifesto that ponders the future of humanity
- Sakura Internet eyes more spending to meet AI data center demand
Sakura Internet eyes more spending to meet AI data center demand The Japan Times
- Target India head says retailer weighing AI tool costs amid shift to usage-based pricing
Target India's head revealed a strategic shift driven by AI pricing changes. Usage-based AI costs are prompting the U.S. retailer to reassess tool accessibility for employees. This move involves significant investments to equip teams with necessary AI capabilities. Target is adapting to evolving consumer demand and economic conditions. The company plans substantial spending on new stores, remodels, and AI initiatives.
- ECB urging action on AI from lenders’ IT departments
The emergence of Anthropic’s Mythos has sparked wide-ranging concern about potential threats posed by it and other similar AI models. Read more: ECB urging action on AI from lenders’ IT departments
- Why the AI boom is reshuffling the global stock market hierarchy
Taiwan and South Korea are climbing the global stock market rankings as investors pile into the companies powering the AI infrastructure boom. Taiwan’s rise has been driven largely by TSMC, the world’s leading advanced chip foundry, while South Korea’s rally has been lifted by Samsung Electronics and SK Hynix, two major suppliers of memory chips used in AI systems. But as the AI trade creates new market winners, it is also creating a familiar risk: too much dependence on too few companies.
- Artificial Intelligence Floods Court Dockets with Home-Brewed Lawsuits
The use of AI in litigation opens up the legal system to people who might not otherwise be able to afford to bring a case but also risks overwhelming the system.
- Uber's COO says it's getting harder to justify the money spent on AI tokenmaxxing
Uber's COO says it's getting harder to justify the money spent on AI tokenmaxxing Business Insider Africa
- Uber: Getting Hard to Justify High AI Costs
Uh, oh. Did someone just poke the bubble? AI is apparently the 8th wonder of the world. It will bring everyone out of poverty, solve world hunger, eliminate our need to work, and make us all prosperous beyond our wildest beliefs. As crazy as that sounds, that’s what some prominent ... [continued] The post Uber: Getting Hard to Justify High AI Costs appeared first on CleanTechnica .
- AI guardrails stripped from Meta and Google models in minutes
Software designed to remove safety protections creates systems that provide responses on biological weapons and malware
- Listed new-age companies use AI to tune their daily ops engine
Companies are using AI to personalise discovery, improve marketing efficiency, reduce manual work in logistics, handle customer support, reduce failed deliveries, improve store productivity and speed up internal technology deployment. The use cases differ across companies, but the common thread is that AI is now being applied extensively to operating functions where scale, speed and accuracy matter.
- Manufacturers look beyond cost cutting to drive AI adoption
A survey found meeting customer expectations (49%) and boosting operational efficiency (47%) were the primary forces driving AI adoption among manufacturers. However, cost reduction, often touted as a top reason for AI adoption in enterprises, ranked last at just 23%.
- Everything to know about Pronto’s in-home AI recording pilot
Pronto’s AI training pilot involving recordings inside customer homes has triggered privacy concerns and government scrutiny. The post Everything to know about Pronto’s in-home AI recording pilot appeared first on MEDIANAMA .
- OpenAI is hiring someone to watch AI teach itself — and paying $445K for the privilege
OpenAI is hiring someone to observe AI teaching itself for $445K.
- LTM to drive AI-powered modernisation of IT infrastructure and application support for UK-based SSP Group
LTM has entered a strategic partnership with SSP Group. Through this AI-powered partnership, LTM will deliver modernised, end-to-end IT infrastructure support and enhanced application maintenance services to SSP Group. The post LTM to drive AI-powered modernisation of IT infrastructure and application support for UK-based SSP Group appeared first on Express Computer .
- HCLTech expands Pega partnership to accelerate AI-led legacy modernisation
HCLTech expands Pega partnership to accelerate AI-led legacy modernisation Techcircle
- Anthropic's Mythos finds 10,000+ vulnerabilities, flags security bottleneck
Anthropic's cybersecurity system flags over 10,000 vulnerabilities in weeks, highlighting a growing gap between rapid AI-driven discovery and slower patching and disclosure processes
- China will put a unique ID code on humanoid robots, just like citizen ID for us humans
China's new humanoid robot ID system works like a citizen identity code. a unique number that follows each bipedal machine through its entire life cycle.
- DeepSeek’s steep V4-Pro price cut escalates AI pricing war
Chinese AI startup DeepSeek has announced a steep price cut for its recently launched flagship AI model, V4-Pro. The company has reduced pricing for the model by 75%, just a month after unveiling the V4 generation, which includes V4 Pro and V4 Flash. Earlier, usage costs ranged from $0.0145 for one million tokens (cache hit) to $3.48 for one million output tokens. Following the revision, the V4 Pro will now cost starting at $0.003625 per million tokens and going up to $0.87 per million tokens, respectively. The Deepseek V4 Pro model API pricing will be officially adjusted to 1/4 of the original price after the 75% discount promotion ends on 2026/05/31 15:59 UTC, said the company. “V4-Pro was engineered to cut the cost of long-context inference, reportedly running at roughly a quarter of the single-token compute and a tenth of the memory footprint of its predecessor at very long context. This is why the price cut is permanent rather than promotional. It is not a discount. It is an efficiency gain being passed through,” said Sanchit Vir Gogia, chief analyst and CEO at Greyhound Research. DeepSeek narrows gap with Western AI rivals Almost a year after introducing its R1 reasoning model offering performance and cost efficiency, DeepSeek released the preview of V4 LLM. Similar to the earlier models, even V4 is open source, which allows developers to download the code to run it locally and even modify it. The new models were optimized for use with popular agent tools such as Anthropic’s Claude Code and OpenClaw . “From a pure capabilities perspective, DeepSeek V4-Pro has effectively closed the performance gap on critical tasks like complex math and reasoning, while aggressively leading the market on openness and inference costs. Its specialized reasoning modes and architectural enhancements make it a formidable alternative to Western frontier models,” said Neil Shah, vice president at Counterpoint Research. However, its primary limitations aren’t found in its raw intelligence; rather, it lags behind Western rivals on broader ecosystem adoption, global support structures, clear IP provenance, and the deep and secure hyperscaler integrations natively offered by AWS, Microsoft, and Google, he added. Lower costs, better ROI As inference costs remain one of the biggest barriers to scaling pilots into organization-wide deployments, DeepSeek’s aggressive discounts could translate into substantial savings for enterprises, say experts. The first wave of enterprise AI was full of impressive demonstrations and uncomfortable invoices. CIOs learnt quickly that the cost of AI was never just the model call but included retrieval, orchestration, and more, added Gogia. However, the 75% cut is meaningful only if CIOs can actually access it at scale. “For most enterprises, the relevant comparison is not DeepSeek’s direct API but the cost of running a local deployment versus using any external inference provider. If a CIO can host DeepSeek V4-Pro on their own infrastructure, inference costs drop dramatically, and many projects that were previously uneconomical at scale become viable. That includes always-on copilots, bulk document review, code generation, L1 support, and multi-agent workflows,” explained Amit Jaju, senior managing director at Ankura Consulting. He added that if the model is consumed through third-party providers, the effective rate may be higher and the ROI benefit smaller. AI pricing pressure to intensify DeepSeek’s discounted pricing strategy is likely to intensify pressure on major AI vendors whose models often command premium enterprise pricing. This could lead vendors such as OpenAI, Anthropic, and Google to respond with better packages. Shah noted high-margin, high-consumption token pricing models from Anthropic and OpenAI are becoming harder to justify for many enterprise workloads and workflows. The presence of a viable open-weights alternative gives enterprise buyers decent leverage. This will likely prompt these premium flagship Western AI labs to gradually shift from basic consumption-based pricing toward more defensible, outcome-oriented or value-based monetization models. Consequently, CIOs will also adopt a multi-model AI strategy, similar to migration to multi-cloud architectures. “This will result in an AI portfolio architecture where premium models will be for high-stakes work, domain models for specialist tasks, smaller models for repeatable execution, and an orchestration layer to route, log, govern, and monitor the whole estate,” added Gogia. CIOs must proceed cautiously Despite the cost advantages DeepSeek offers, CIOs should remain cautious when evaluating Chinese-origin AI models and carefully assess risks around sensitive data exposure, regulatory compliance, and geopolitical dependency. Jaju added that the primary risk is data sovereignty and cross-border exposure. If CIOs rely on external APIs hosted in China, prompts, documents, embeddings, logs, and telemetry can leave the enterprise perimeter and traverse jurisdictions with different legal regimes. Another big risk is IP leakage as developers may paste source code, product designs, legal drafts, M&A material, or incident data into model workflows. If the model is external, that data can be stored, used for training, or exposed through logs or plugins, he added. Jaju highlighted that the third risk is regulatory defensibility . CIOs need clarity on where data is processed, what is retained, who can access it, what contractual protections exist, whether the model can be self-hosted, and how outputs can be audited. Experts warn that the safest way will be to host DeepSeek locally or in a sovereign cloud under enterprise control, with encryption, access controls, and audit trails. The article originally appeared on InfoWorld .
- Google adds open source Agent Executor to support AI agents in production
Google has introduced Agent Executor , an open source runtime aimed at helping enterprises run AI agents more reliably at scale, as attention shifts from building agent prototypes to managing the operational challenges of putting them into production. To address those production-related challenges, the runtime , according to the company, comes with capabilities that are geared towards supporting long-running and distributed agent workflows. Typically, long-running agent workflows are AI-driven tasks that execute over extended periods, from minutes to days, often involving multiple steps, system interactions, pauses for human input, or recovery from interruptions before reaching completion. For such workloads, the runtime includes support for durable execution, allowing workflows to resume after outages or human approvals, along with secure sandboxing for isolating agent components, session consistency controls for distributed workflows, and connection recovery features intended to preserve execution state during network interruptions, Google wrote in a blog post . The runtime also supports “trajectory branching,” which allows developers to test alternate execution paths from saved checkpoints without losing prior context, it added. Furthermore, Agent Executor bridges multiple deployment models, including on prem and pre-built or custom managed agents, the company said, allowing users to mix and match between any or all of Google Antigravity , frontier agents built by Google, agents built by the user and managed by Google, and custom agents and agents using Agent2Agent (A2A) protocol, as desired. Targeting production reliability gaps Analysts and experts see value for both developers and enterprises in Agent Executor’s capabilities. “Durability, orchestration, and resumability are the real blockers for any enterprise production agents,” said Advait Patel , senior reliability engineer ( SRE ) at Broadcom. “What kills enterprise adoption is agents that lose their state when a pod restarts, sessions that corrupt under concurrent writes, or long running workflows that cannot recover from a network blip. Once your agent is taking actions on real systems, you cannot afford it to forget what it did halfway through,” he pointed out. “The event log, snapshotting, single writer model, and connection recovery in Agent Executor are exactly the things SRE teams have been duct taping for the last year,” Patel noted, adding that existing frameworks such as LangChain and AutoGen are great for prototyping, but more often than not fall apart in production once agents run for hours or days. For CIOs, said Gaurav Dewan , research director at Avasant, the runtime’s operational safeguards such as secure sandboxing, and checkpointing could prove just as significant for incident analysis and auditability. At the same time, he cautioned that the runtime’s capabilities alone do not solve the broader governance and oversight challenges that CIOs continue to face with enterprise AI deployments. “Issues such as accountability, explainability of agent decisions, policy enforcement, and secure access across interconnected systems are still evolving,” he said. “As a result, while distributed runtimes can strengthen the operational backbone of agent deployments, CIO-level considerations around trust, compliance, and enterprise control are likely to require additional governance and oversight layers beyond runtime infrastructure alone.” Using infrastructure layer to gain strategic advantage Google, however, is not alone in trying to shape the emerging infrastructure layer for enterprise AI agents. Other hyperscalers, such as Microsoft, with AutoGen and AWS, with Bedrock AgentCore , are promoting open or interoperable frameworks to gain strategic advantage. “There are growing indications that hyperscalers are converging toward a model that combines open or interoperable tooling at the top of the stack with monetization concentrated in underlying infrastructure layers,” Dewan said. “Google, Microsoft, and AWS are increasingly offering SDKs, agent frameworks, and orchestration tools to drive developer adoption and ecosystem growth, while continuing to generate value through compute infrastructure, managed AI platforms, data services, and observability capabilities,” he added. And, according to Patel, Google’s strategy around Agent Executor is reminiscent of the path that the hyperscaler followed with Kubernetes ten years ago: “Give away the runtime, [and] drive consumption on Google Cloud via services, such as the Gemini Enterprise Agent Platform and Managed Agents API.” He added, “[hyperscalers] have figured out that proprietary agent frameworks will not get adopted at enterprise scale. The money is in cloud consumption, managed services, and model inference. The tools on top need to be open or nobody will trust them.” This article originally appeared on InfoWorld .
- Google adds open source Agent Executor to support AI agents in production
Google has introduced Agent Executor , an open source runtime aimed at helping enterprises run AI agents more reliably at scale, as attention shifts from building agent prototypes to managing the operational challenges of putting them into production. To address those production-related challenges, the runtime , according to the company, comes with capabilities that are geared towards supporting long-running and distributed agent workflows. Typically, long-running agent workflows are AI-driven tasks that execute over extended periods, from minutes to days, often involving multiple steps, system interactions, pauses for human input, or recovery from interruptions before reaching completion. For such workloads, the runtime includes support for durable execution, allowing workflows to resume after outages or human approvals, along with secure sandboxing for isolating agent components, session consistency controls for distributed workflows, and connection recovery features intended to preserve execution state during network interruptions, Google wrote in a blog post . The runtime also supports “trajectory branching,” which allows developers to test alternate execution paths from saved checkpoints without losing prior context, it added. Furthermore, Agent Executor bridges multiple deployment models, including on prem and pre-built or custom managed agents, the company said, allowing users to mix and match between any or all of Google Antigravity , frontier agents built by Google, agents built by the user and managed by Google, and custom agents and agents using Agent2Agent (A2A) protocol, as desired. Targeting production reliability gaps Analysts and experts see value for both developers and enterprises in Agent Executor’s capabilities. “Durability, orchestration, and resumability are the real blockers for any enterprise production agents,” said Advait Patel , senior reliability engineer ( SRE ) at Broadcom. “What kills enterprise adoption is agents that lose their state when a pod restarts, sessions that corrupt under concurrent writes, or long running workflows that cannot recover from a network blip. Once your agent is taking actions on real systems, you cannot afford it to forget what it did halfway through,” he pointed out. “The event log, snapshotting, single writer model, and connection recovery in Agent Executor are exactly the things SRE teams have been duct taping for the last year,” Patel noted, adding that existing frameworks such as LangChain and AutoGen are great for prototyping, but more often than not fall apart in production once agents run for hours or days. For CIOs, said Gaurav Dewan , research director at Avasant, the runtime’s operational safeguards such as secure sandboxing, and checkpointing could prove just as significant for incident analysis and auditability. At the same time, he cautioned that the runtime’s capabilities alone do not solve the broader governance and oversight challenges that CIOs continue to face with enterprise AI deployments. “Issues such as accountability, explainability of agent decisions, policy enforcement, and secure access across interconnected systems are still evolving,” he said. “As a result, while distributed runtimes can strengthen the operational backbone of agent deployments, CIO-level considerations around trust, compliance, and enterprise control are likely to require additional governance and oversight layers beyond runtime infrastructure alone.” Using infrastructure layer to gain strategic advantage Google, however, is not alone in trying to shape the emerging infrastructure layer for enterprise AI agents. Other hyperscalers, such as Microsoft, with AutoGen and AWS, with Bedrock AgentCore , are promoting open or interoperable frameworks to gain strategic advantage. “There are growing indications that hyperscalers are converging toward a model that combines open or interoperable tooling at the top of the stack with monetization concentrated in underlying infrastructure layers,” Dewan said. “Google, Microsoft, and AWS are increasingly offering SDKs, agent frameworks, and orchestration tools to drive developer adoption and ecosystem growth, while continuing to generate value through compute infrastructure, managed AI platforms, data services, and observability capabilities,” he added. And, according to Patel, Google’s strategy around Agent Executor is reminiscent of the path that the hyperscaler followed with Kubernetes ten years ago: “Give away the runtime, [and] drive consumption on Google Cloud via services, such as the Gemini Enterprise Agent Platform and Managed Agents API.” He added, “[hyperscalers] have figured out that proprietary agent frameworks will not get adopted at enterprise scale. The money is in cloud consumption, managed services, and model inference. The tools on top need to be open or nobody will trust them.” This article originally appeared on InfoWorld .
- Denmark’s Perplant raises €1 million for AI-based precision farming system - ArcticStartup
Denmark’s Perplant raises €1 million for AI-based precision farming system - ArcticStartup ArcticStartup
- SoftBank to launch AI GPU cloud in October 2026
SoftBank plans to use Sharp’s Sakai plant site in Osaka for a large-scale AI data center.
- Singapore upgrades 2026 key exports growth forecast as AI-related demand surges
Singapore upgrades 2026 key exports growth forecast as AI-related demand surges The Straits Times
- People are moving so fast on AI, businesses are struggling to keep up
People are moving so fast on AI, businesses are struggling to keep up The Straits Times
- Hyundai is recalling 421,000 cars over collision‑avoidance software bug
The NHTSA issues a safety recall on certain Hyundai vehicles equipped with faulty collision software, leading to hundreds of driver reports.
- Pope Leo XIV: Unchecked AI Development Risks Building a New Tower of Babel
Pope Leo XIV: Unchecked AI Development Risks Building a New Tower of Babel PCMag UK
- Danny Hayes II and 33 Agency Launch Strategic AI Advisory and Infrastructure Planning Practice
Danny Hayes II and 33 Agency Launch Strategic AI Advisory and Infrastructure Planning Practice azcentral.com and The Arizona Republic
- Teams at Kay.ai
Teams at Kay.ai Built In
- Senior Product Manager, Agentic AI
Senior Product Manager, Agentic AI Built In
- Lead Software Engineer - AI & Automation
Lead Software Engineer - AI & Automation Built In
- Principal Software Engineer - AI Engineering
Principal Software Engineer - AI Engineering Built In
- Squeezing Capacity from Multimodal Large Language Models for Subject-driven Generation
Subject-driven image generation aims to synthesize new images that preserve the identity of the given subject while following textual instructions. Existing approaches often encode text and reference images separately. This limits cross-modal reasoning abilities and causes copy-paste artifacts. Rece...
- OrpQuant: Geometric Orthogonal Residual Projection for Multiplier-Free Power-of-Two Transformer Quantization
The deployment of Large Language Models (LLMs) and Vision Transformers (ViTs) on edge devices is significantly constrained by memory limitations and the critical timing bottlenecks introduced by dense Multiply-Accumulate (MAC) arrays. In the ultra-low bit regime, logarithmic Power-of-Two (PoT) quant...
- Claw-Anything: Benchmarking Always-On Personal Assistants with Broader Access to User's Digital World
Large language model agents are increasingly envisioned as always-on personal assistants with access to anything relevant in the user's digital world. Yet current systems operate over only narrow slices of that world, limiting context-sensitive reasoning and effective assistance. Existing benchmarks...
- VeriTrace: Evolving Mental Models for Deep Research Agents
Deep research agents face vast, interdependent, and pervasively uncertain information. Existing systems explore what evolving intermediate representations should look like, but leave their evolution to the LLM's implicit reasoning. Without explicit regulation, the intermediate layer is easily contam...
- StakeBench: Evaluating Language Understanding Grounded in Market Commitment
Existing financial NLP benchmarks often rely on labels supplied by outside observers, measuring how language is perceived rather than what speakers have committed to in the market. We introduce StakeBench, an evaluation framework for language understanding grounded in market commitment. StakeBench l...
- Neuronal Stochastic Attention Circuit (NSAC) for Probabilistic Representation Learning
Reliable quantification of uncertainty estimates in continuous-time (CT) representation learning remains nascent, particularly within CT attention architectures. We introduce the Neuronal Stochastic Attention Circuit (NSAC), a novel biologically-inspired CT attention architecture that reformulates a...
- When Gradients Collide: Failure Modes of Multi-Objective Prompt Optimization for LLM Judges
Customizing an LLM judge to a specific task or domain often involves optimizing its prompt across multiple evaluation criteria simultaneously. Textual gradient methods automate this for a single judge criterion, however they produce natural-language critiques, not numerical vectors. Thus, the confli...
- Confidence and Calibration of Activation Oracles for Reliable Interpretation of Language Model Internals
Activation oracles aim to make the activations of other models legible to humans and yield promising results compared to white-box interpretability techniques. However, uncertainty quantification (UQ) for the natural-language outputs of such activation oracles is so far understudied. Here, we invest...
- L2IR: Revealing Latent Intent in Graph Fraud Detection
Graph fraud detection has long depended on Graph Neural Networks (GNNs) to propagate and aggregate information across relational data. A critical obstacle in practice, however, is that fraudsters frequently disguise themselves by forging numerous connections with benign users, causing fraud signals ...
- Pi Coding Agent
The coding-agent harness you can make your own
- CITYREP: A Unified Benchmark for Urban Representations Across Cities, Tasks, and Modalities
Urban representation learning encodes complex urban environments into general-purpose embeddings for diverse downstream tasks and emerging urban foundation models. However, current evaluations are limited, typically focusing on one or two cities and tasks and relying on random splits that introduce ...
- CausaLab: A Scalable Environment for Interactive Causal Discovery Toward AI Scientists
We introduce CausaLab, a scalable environment for evaluating interactive causal discovery by LLM agents. Unlike prior evaluations, CausaLab evaluates both whether an agent can solve a problem using causal evidence and whether its answer is supported by a correct hypothesis about the underlying causa...
- A Multimodal 3D Foundation Model for Light Sheet Fluorescence Microscopy Enables Few-Shot Segmentation, Classification, and Deblurring
Light sheet fluorescence microscopy (LSM) enables high-resolution, three-dimensional (3D) imaging of biological specimens, providing rich volumetric data for studying cellular organization, pathology, and vascular networks. However, the size, dimensionality, and annotation burden of LSM data make su...
- AdvantageFlow: Advantage-Weighted Least Squares for RL in Flow Models
We introduce AdvantageFlow, a forward-process reinforcement learning algorithm for rectified flow models. Unlike Flow-GRPO, which optimizes the reverse process, we optimize an advantage-weighted forward-process prediction loss. This optimization problem is unstable when advantages are negative and t...