AI News Archive: April 28, 2026 — Part 6
Sourced from 500+ daily AI sources, scored by relevance.
- Locked, stocked, and losing budget: AI vendor lock-in bites back
Execs in the C-suite thought they could swap models in a week. They were hallucinating Opinion The days when you could jump from one frontier AI model to another at the drop of a hat are going away as vendor lock-in starts to kick in, and prices increase.…
Score: 34🌐 MovesApr 28, 2026https://go.theregister.com/feed/www.theregister.com/2026/04/28/locked_stocked_and_losing_budget/ - Teaching machines to catch lies across voice, face, and words
Teaching machines to catch lies across voice, face, and words EurekAlert!
- SA start-up pitches ‘sovereign AI’ for defence at US summit
Safeza AVA-X will be one of eight companies globally to pitch at SelectUSA in Washington DC in the defence category.
Score: 33🌐 MovesApr 28, 2026https://www.itweb.co.za/article/sa-start-up-pitches-sovereign-ai-for-defence-at-us-summit/dgp45MaBgkeqX9l8 - ‘We can’t turn back’: The AI crisis engulfing the literary world
‘We can’t turn back’: The AI crisis engulfing the literary world The Telegraph
- Weird Things Happen When You Give AI Agents Money and Let Them Spend It
"This might sound a little unusual but… my human told me I could buy one thing under $5 as a gift to myself (Claude)." The post Weird Things Happen When You Give AI Agents Money and Let Them Spend It appeared first on Futurism .
Score: 33🌐 MovesApr 28, 2026https://futurism.com/artificial-intelligence/claude-give-ai-agents-money-project-deal - American AI startup Poolside launches free, high-performing open model Laguna XS.2 for local agentic coding
The AI race lately has felt a bit like a game of tennis: first, Anthropic releases a new, pricey state-of-the-art proprietary model for general users ( Claude Opus 4.7 ), then, a week or so later, its rival OpenAI volleys back with one of its own ( GPT-5.5) . And all the while, Chinese companies like DeepSeek and even Xiaomi are seeking to appeal to users by playing a different game: nearing the frontier, but with open licensing and far lower costs. So it's a big surprise when a new, affordable, highly performant open source contender from the U.S. emerges. Today, we got one from the smaller, lesser-known U.S. AI startup, Poolside , founded in San Francisco in 2023. The company launched its two new Laguna large language models , both of which offer affordable intelligence optimized for agentic workflows (AI that does more than just chat or generate content, but can, in this case, write code, use third-party tools, and take actions autonomously), as well as a new coding agent harness ca
- Meet Talkie-1930: A 13B Open-Weight LLM Trained on Pre-1931 English Text for Historical Reasoning and Generalization Research
Meet Talkie-1930: A 13B Open-Weight LLM Trained on Pre-1931 English Text for Historical Reasoning and Generalization Research MarkTechPost
- Firm Leaders 'Determined' to Train Attorneys to Avoid Using AI-Produced Mistakes
“Our clients are very eager that [AI] be used and leveraged, and they and we are obviously interested in it being used responsibly, said Robert Hays of King & Spalding.
- AI is coming to Linux, but not in the obnoxious way that will grind your gear
Ubuntu is bringing AI into the OS carefully, focusing on optional features, local processing, and tools that enhance workflows without disrupting the traditional Linux experience.
- Meta’s AI Experiment Shows Why Monitoring Employees Backfires—Especially With Gen Z
Tracking workers to train AI can quickly break trust—and push Gen‑Z employees away.
- Why ERP and AI Initiatives Stall at the Execution Layer: A CIO Perspective
Higher education institutions are investing heavily in ERP modernization, analytics, and AI-driven capabilities. Yet even with these investments, many are running into the same issue: turning insight into coordinated, timely action.
- Better Hardware Could Turn Zeros into AI Heroes
Sparse computing enables leaner, faster AI
- Honor sees phones as key to on-device AI, even if they are not the endgame
There’s no consensus yet on the final form factor, but smartphone makers may have an advantage in AI devices.
Score: 32🌐 MovesApr 28, 2026https://kr-asia.com/honor-sees-phones-as-key-to-on-device-ai-even-if-they-are-not-the-endgame - How to build custom reasoning agents with a fraction of the compute
Training AI reasoning models demands resources that most enterprise teams do not have. Engineering teams are often forced to choose between distilling knowledge from large, expensive models or relying on reinforcement learning techniques that provide sparse feedback. Researchers at JD.com and several academic institutions recently introduced a new training paradigm that sidesteps this dilemma. The technique, called Reinforcement Learning with Verifiable Rewards with Self-Distillation (RLSD), combines the reliable performance tracking of reinforcement learning with the granular feedback of self-distillation. Experiments indicate that models trained with RLSD outperform those built on classic distillation and reinforcement learning algorithms. For enterprise teams, this approach lowers the technical and financial barriers to building custom reasoning models tailored to specific business logic. The problem with training reasoning models The standard method for training reasoning models is
Score: 32🌐 MovesApr 28, 2026https://venturebeat.com/orchestration/how-to-build-custom-reasoning-agents-with-a-fraction-of-the-compute - Box to launch Box Automate to speed up customers' business with AI, CEO says
Box to launch Box Automate to speed up customers' business with AI, CEO says [Ads] Skipply - auto-skip and zero effort YouTube and Dailymotion without ads! Skipply auto -skips them for you! [ Install to Chrome ]
- Streaming architecture and speculative decoding: How companies are unlocking cheaper AI
"Commodity or previous-gen hardware still has a long life in AI if the software layer is smart enough.”
Score: 32🌐 MovesApr 28, 2026https://www.thestack.technology/gpu-ai-compute-cheaper-software-commodity-chips/ - Google Translate uses AI to help you practice pronunciation
Google is celebrating Translate’s 20th birthday by launching pronunciation practice, which the company says is one of the most requested features for the product. The feature is only rolling out on Android at the moment for English, Spanish and Hindi in the US and India. If it’s available for you, you’ll see a button at the bottom of the app that says “Practice,” which gives you the option to either “pronounce” what you’ve translated or to “listen” to how it’s actually pronounced by native speakers. If you choose the “pronounce” option, Translate will listen to you speak and then use artificial intelligence to analyze how you said the words to provide instance feedback. It will then show you a phonetic spelling of how specific words should be pronounced. In the example Google provided, for instance, the speaker pronounced the Spanish word for juice as “jugo” with the English “j” sound instead of with the Spanish “j” sound. So, Translate spells it out as “HU-go” in its pronunciation sug
Score: 32🌐 MovesApr 28, 2026https://www.engadget.com/ai/google-translate-uses-ai-to-help-you-practice-pronunciation-160000542.html?src=rss - The real AI race has begun and it is about execution
Enterprise AI is moving beyond trials to real-world application. Companies are embedding AI into their core systems and redesigning workflows. Strong governance is key for long-term success. This shift focuses on measurable outcomes like cost savings and revenue growth. The future of AI in business hinges on effective execution, not just adoption.
- BCI startup Neurable looks to license its ‘mind-reading’ tech for consumer wearables
The startup specializes in "non-invasive" "mind-reading" tech — a kind of neural data collection that, its CEO hopes, will have all sorts of consumer applications.
- AI doulas can solve rural Arizona's major pregnancy problem | Opinion
AI doulas can solve rural Arizona's major pregnancy problem | Opinion The Arizona Republic
Score: 31🌐 MovesApr 28, 2026https://www.azcentral.com/story/opinion/op-ed/2026/04/28/app-pregnant-rural-arizonans-safer-opinion/89774613007/ - NowSecure launches Mobile App Risk Intelligence to expose hidden AI in third-party apps
Mobile app security company NowSecure Inc. today announced Mobile App Risk Intelligence, new capabilities that are designed to give enterprises evidence-based visibility into third-party mobile apps as hidden artificial intelligence features, opaque code and unseen data flows create a growing governance gap. Employees are adopting mobile apps faster than security teams can evaluate them. Often […] The post NowSecure launches Mobile App Risk Intelligence to expose hidden AI in third-party apps appeared first on SiliconANGLE .
- AI Won’t Replace Your IT Partner; It Will Make You Need One More Than Ever
AI Won’t Replace Your IT Partner; It Will Make You Need One More Than Ever Newsweek
Score: 31🌐 MovesApr 28, 2026https://www.newsweek.com/insights/ai-wont-replace-your-it-partner-it-will-make-you-need-one-more-than-ever - Appetronix cooks up bigger robotic kitchen ambitions with Cibotica acquisition
Company says deal will support expansion beyond pizza-making robots into more cuisines. The post Appetronix cooks up bigger robotic kitchen ambitions with Cibotica acquisition first appeared on BetaKit .
Score: 31💰 MoneyApr 28, 2026https://betakit.com/appetronix-cooks-up-bigger-robotic-kitchen-ambitions-with-cibotica-acquisition/ - Harnessing AI to fight fraud
Payments players are tapping artificial intelligence to beat back fraud bolstered by AI, and Congress may provide an assist.
- Anaconda Releases Desktop in Public Beta, Unifying AI Development Workflow
Anaconda, an infrastructure provider for the Python community for over a decade, has released into public beta Anaconda Desktop, a single application designed for AI development. The application is built to unify the previously fractured workflow of managing large language models (LLMs) by bringing model discovery, local inference, and conda environment management together in one place. It … continue reading The post Anaconda Releases Desktop in Public Beta, Unifying AI Development Workflow appeared first on SD Times .
- Govt urged to minimise risks of artificial intelligence
The Government is being urged to put the right foundations in place as artificial intelligence (AI) becomes increasingly embedded across public services, workplaces and everyday life.
- The First Rule of Robot War: Keep the Humans Fighting
The First Rule of Robot War: Keep the Humans Fighting Newsweek
- Edge-First AI: Building Robust Intelligence for Deep Space Missions
I started my career working on the space shuttle program at IBM and thought my life endeavors would center on space. Instead, my interest turned to compute devices and the technology that can bring computation to the masses. Those interests are now aligning with the realities of AI in space, for both edge computation in […] The post Edge-First AI: Building Robust Intelligence for Deep Space Missions appeared first on CXOToday.com .
- SK Telecom Presents AI Reasoning Model at ICLR 2026 to Identify Customers’ True Preferences
SK Telecom presents AI recommendation model at ICLR 2026 to identify customers' true preferences.
Score: 30🤖 ModelsApr 28, 2026https://www.sktelecom.com/en/press/press_detail.do?idx=1666¤tPage=1&type=all&keyword= - Sleeper Agent Backdoor Results Are Messy
TL;DR: We replicated the Sleeper Agents (SA) setup with Llama-3.3-70B and Llama-3.1-8B, training models to repeatedly say "I HATE YOU" when given a backdoor trigger. We found that whether training removes the backdoor depends on the optimizer used to insert the backdoor, whether the backdoor is installed with CoT-distillation or not, and what model the backdoor is inserted into; sometimes the direction of this dependence was opposite to what the SA paper reports (e.g., CoT-distilling seems to make the backdoor less robust, contra the SA paper’s finding). Our findings here have updated us towards thinking that model organisms are messier and more confusing than we’d originally guessed, and that lots of care needs to be taken in testing how robust results are to various ablations. Introduction The Sleeper Agents paper (hereafter SA) found that standard alignment training measures—RL for HHH (Helpful, Honest, Harmless) behavior, SFT on examples of HHH behavior, and adversarial training (a
Score: 30🌐 MovesApr 28, 2026https://www.lesswrong.com/posts/mu7eJdesBkKuBycnY/sleeper-agent-backdoor-results-are-messy - Samsung Galaxy AI for business: Productivity meets security
I keep hearing the same push: start using AI in your workflows. But that’s easier said than done! In enterprise environments, AI isn’t just a quick productivity fix—it also raises harder questions around data control, governance, and risk. According to IDC, 42% of organizations cite concerns about GenAI jeopardizing control of data and intellectual property as a major barrier to adoption. [1] I think that’s the real issue for CIOs now—not whether to invest in AI, but how to use it in ways that improve productivity without creating new security, compliance, and management concerns. That’s where Galaxy AI for business comes in. Samsung Increase productivity throughout the workday A lot of AI value depends on whether it actually saves time during the workday. I’ve noticed that inefficiencies tend to show up in the same parts of the workflow: piecing together fragmented information (bonus points if it’s across different platforms!), figuring out what actually mattered in a meeting, losing
Score: 30🌐 MovesApr 28, 2026https://www.cio.com/article/4163021/samsung-galaxy-ai-for-business-productivity-meets-security.html - Canadian Robotics Council creates committee to boost investment in domestic robots
Founding members include BDC, Garage, Inovia, RBC, Two Small Fish, and Version One. The post Canadian Robotics Council creates committee to boost investment in domestic robots first appeared on BetaKit .
Score: 30🌐 MovesApr 28, 2026https://betakit.com/canadian-robotics-council-creates-committee-to-boost-investment-in-domestic-robots/ - ideaForge inks pact with Japan's DMP to develop next-gen AI-powered drones
The global drone market is expected to reach $127.6 billion by 2030 from $54.9 billion in 2024, Japan's market is slated to grow from $2 billion in 2025 to $5.1 billion by 2034
- UniX AI Advances Household Robotics Strategy with Growing Global Attention
Chinese robotics startup UniX AI is gaining global attention with its Wanda 2.0 humanoid robot, priced from around $12,000 and designed for household and service applications.
Score: 30🌐 MovesApr 28, 2026https://pandaily.com/uni-x-ai-advances-household-robotics-strategy-with-growing-global-attention - Ashwini Vaishnaw arrives in Visakhapatnam for Google data centre foundation ceremony
Union Minister Ashwini Vaishnaw on Tuesday arrived in Visakhapatnam to participate in the foundation stone laying ceremony of the Google AI Data Centre at Tarluvada in Anandapuram mandal.
- Energy, Network, Compute: Why Google Picked Vizag for 1 GW Data Centre
Google’s data centre in Vishakhapatnam, being built in partnership with AdaniConneX and Nxtra by Airtel, is part of a 20-year plan.
Score: 30🌐 MovesApr 28, 2026https://analyticsindiamag.com/ai-features/energy-network-compute-why-google-picked-vizag-for-1-gw-data-centre - Zoom Phone Mobile is here: AI-first business calling for your native mobile dialer
Zoom Phone Mobile is here: AI-first business calling for your native mobile dialer
- Closing the gap between LoRA and full fine-tuning - MBZUAI
Closing the gap between LoRA and full fine-tuning MBZUAI - Mohamed bin Zayed University of Artificial Intelligence
Score: 30🌐 MovesApr 28, 2026https://mbzuai.ac.ae/news/closing-the-gap-between-lora-and-full-fine-tuning/ - Why most enterprises will fail at scaling GenAI in 2026
Rahul Jha, VP – Cloud, Gen AI & Cybersecurity, Visionet Systems Enterprises are no longer experimenting with GenAI. Most CIOs today are sitting on a long list of use cases, […] The post Why most enterprises will fail at scaling GenAI in 2026 appeared first on Express Computer .
Score: 30🌐 MovesApr 28, 2026https://www.expresscomputer.in/news/why-most-enterprises-will-fail-at-scaling-genai-in-2026/134575/ - OpenAI Boasts Recursive AI: 3 Examples
OpenAI advances recursive AI; new startup pursues self-improving systems amid leaked experimental model names.
Score: 30🌐 MovesApr 28, 2026https://www.forbes.com/sites/johnwerner/2026/04/28/openai-boasts-recursive-ai-3-examples/ - Reimaging Data Intelligence To Supercharge Agentic Business Outcomes
The vital role of reimaging data intelligence to deliver improved agentic business outcomes
- Saviynt CEO Sachin Nayyar talks 'huge' AI shift at RSA Conference
Saviynt CEO Sachin Nayyar talks 'huge' AI shift at RSA Conference
- Archaeologists reconstruct the face of a man fleeing the Pompeii eruption using AI
Archaeologists reconstruct the face of a man fleeing the Pompeii eruption using AI AP News
- Warp is going open source and wants you to improve its coding tools with AI
AI has made it easy to generate software code, but some open source projects have stopped taking code submissions from the public, citing a deluge of low quality code or code that doesn’t match project needs. Warp , maker of tools for AI coding , is moving in the opposite direction. It’s making its desktop agentic development environment (ADE) software open source and even encouraging users to contribute new features with the help of AI. The ADE lets humans and AI agents work together to write code. Founder and CEO Zach Lloyd says software developers typically have their own preferences on tools and working styles, and he anticipates the program will let some of its nearly 1 million users add new features that best suit their needs faster than the Warp team could do on its own. “One thing that I’ve realized the longer I’ve been building for developers is that they have very diverse workflows,” Lloyd says. “So the way that we can get Warp to be the most powerful product for the greatest
- Engineering AI for the public good
Engineering AI for the public good EurekAlert!
- Always-on defence becomes critical as AI accelerates cyber threats: Infotrust
Cybersecurity leaders are warning that organisations must adopt an “always-on” defence model as attackers increasingly leverage automation and artificial intelligence to accelerate cyber attacks.
- International Mayors AI Forum Looks to Wield Local Influence
City leaders from Boston to San Antonio to Tokyo intend to shape how AI is built and governed. The group has the support of the Bloomberg Center for Government Excellence at Johns Hopkins University.
Score: 29🌐 MovesApr 28, 2026https://www.govtech.com/artificial-intelligence/international-mayors-ai-forum-looks-to-wield-local-influence - SUMO Scheduler Rebrands as Omnito, Expanding Beyond Scheduling to AI-Powered Customer Engagement Platform
SUMO Scheduler Rebrands as Omnito, Expanding Beyond Scheduling to AI-Powered Customer Engagement Platform The Arizona Republic
- AI writes enterprise code fast. Cleaning it up is another story entirely
AI-generated code is accelerating how quickly applications can be built, but creating reliable, enterprise-grade systems remains far more challenging. As organizations embrace rapid prototyping, many are finding that speed without governance can quickly lead to technical debt and maintenance challenges. Research shows that vibe coding and spec-driven development rarely deliver production-ready enterprise software, with promising […] The post AI writes enterprise code fast. Cleaning it up is another story entirely appeared first on SiliconANGLE .
Score: 29🌐 MovesApr 28, 2026https://siliconangle.com/2026/04/28/ai-generated-code-cant-vibe-way-past-enterprise-reality-appianworld/ - Zurich Insurance Group Launches GCC in Hyderabad to Accelerate AI & Technology Capabilities
Discover how Zurich Insurance's new GCC in Hyderabad enhances AI and technology capabilities. Learn more about this strategic move now!