AI News Archive: June 30, 2026 — Part 12
Sourced from 500+ daily AI sources, scored by relevance.
- Vector Institute releases UnBias-Plus, a free, open-source AI tool to detect and rewrite bias in text
Vector Institute releases UnBias-Plus, a free, open-source AI tool to detect and rewrite bias in text Toronto Star
- Microsoft unveils Memora to tackle AI agents’ memory problem
With AI agents increasingly expected to remember conversations, preferences, and decisions over extended periods, Microsoft Research has developed Memora, a memory system designed to provide more scalable and reliable long-term recall than existing approaches. AI agents are increasingly expected to retain context across weeks or months rather than individual chat sessions. Memory can become fragmented, leading to duplicate information and slower retrieval as knowledge grows. According to Microsoft, Memora can solve this problem by decoupling what the AI remembers from how it looks up that information, ultimately reducing context token usage by up to 98% while matching or exceeding full-context accuracy, Microsoft Research claimed in a blog post. Limitations of today’s memory architectures As AI assistants and autonomous agents move into long-horizon deployments, the absence of a principled memory system has become a critical bottleneck. While modern LLMs are powerful reasoners, they still start every session from scratch. Long conversations require models to repeatedly re-read their entire history, while new information is either stored as raw text or compressed into summaries where important details may be lost. Solutions to address these are available, but they too have limitations. For instance, systems like Mem0 extract atomic facts from conversations, retrieval-augmented (RAG) approaches index raw text fragments for later recall, and graph-based memory systems such as Zep and GraphRAG impose structure through entity relations. But these mostly fall into two extremes. Content-fragmentation systems, such as RAG and Mem0, embed extracted facts or text fragments directly. This preserves detail but produces brittle, isolated entries that lose narrative coherence. Coarse-abstraction systems compress experience into compact summaries but strip away the constraints, edge cases, and numeric details that make memory useful in the first place. Graph-based systems add structure on top of content but still rely on the content itself for retrieval and typically require rigid ontologies that don’t generalize across domains. Decoupling memory from retrieval Memora architecture claims to address this by decoupling what is stored from how it is retrieved. For this, each memory entry will have two components. The first will be a primary abstraction, which is a short phrase (6–8 words) that will capture what the memory is fundamentally about. The second will be a memory value, which will hold the rich content itself. As a result of this separation, new information about an evolving topic will be merged into the existing memory entry under the same primary abstraction and will not be fragmented into a chain of partial duplicates. Complementing primary abstractions, cue anchors are short, context-aware tags extracted from each memory’s value, providing alternative access paths to the same memory. They will function as flexible, organically-generated metadata, claimed the post. Memora also introduces a policy-guided retriever that, rather than returning the top-k semantically similar items in a single shot, iteratively refines its query, expands through cue anchors to surface related-but-not-similar memories, and decides when to stop. “The deepest flaw in current agent memory is that it mistakes retrieval for memory. A vector store is superb at finding text that looks relevant. An enterprise agent needs more than resemblance. It needs to know what has changed, what still holds true, and what should never be recalled in the task at hand,” said Sanchit Vir Gogia, chief analyst at Greyhound Research. Memora is interesting precisely because it refuses that shortcut, Gogia noted. It separates the rich detail of a memory from the handle used to find it, indexing a stable abstraction and a set of cue anchors while keeping the full content intact beneath them. Retrieval then becomes an act of navigation rather than a single hopeful guess, as the system re-queries, widens its search, or stops once it has enough, he added. Benchmarking Memora Microsoft evaluated Memora on two long-context benchmarks. LoCoMo, where dialogues average 600 turns, and LongMemEval, which uses 115,000-token contexts. According to the company, Memora achieved 86.3% LLM-judge accuracy on LoCoMo and 87.4% on LongMemEval, outperforming RAG, Mem0, Nemori, Zep, LangMem, and even full-context inference. It also stored nearly half as many memory entries per conversation as Mem0 (344 versus 651) while reducing token consumption by up to 98% compared with full-context inference. While the benchmark results suggest significant efficiency gains, enterprises should not assume lower token consumption will automatically translate into lower infrastructure costs. Gogia cautioned against taking the token reduction number at face value. It is a benchmark context reduction, not a promise that an enterprise bill will fall by 98%, he said. “Real cost also includes memory construction, indexing, storage, and the audit logging that governance demands.” He warned that Memora’s strongest retrieval mode is also its slowest. Its policy retriever runs at between roughly five and six seconds per query across several model-calling steps, against under a second for the simpler semantic mode. The saving in prompt tokens is partly repaid as retrieval latency and extra inference. So the memory crunch does not disappear but moves. Instead of paying only for longer prompts, enterprises must now manage what is written, updated, and forgotten, and the indexing and testing that govern it. Enterprise implications Memora is currently an active Microsoft Research project, but the company has made the research code available on GitHub, enabling developers to experiment with the architecture and adapt it for their own AI applications. However, portability on paper should not be confused with production readiness. While a memory layer of this design can, in principle, sit above models from any major provider, Gogia suggests that until the code is fully verifiable, maintained, and supportable under enterprise controls, the prudent posture for IT leaders is to study Memora as an architecture rather than operationalize it as software. Beyond the technology, organizations will need governance and compliance policies to ensure AI memories are managed securely and remain auditable. He noted an enterprise must decide who may write to memory, who may read it, how long it persists, and how an auditor reconstructs why a memory shaped an action. “An enterprise must decide who may write to memory, who may read it, how long it persists, and how an auditor reconstructs why a memory shaped an action. ‘The agent remembered it’ will not satisfy a regulator under the European Union’s AI Act traceability duties, nor a customer under India’s Digital Personal Data Protection Act,” Gogia said. The article originally appeared on InfoWorld .
- Anthropic to provide California government with AI tools
California government agencies will use Anthropic's AI tools like chatbot Claude for analysing information and drafting documents.
- Gemini Spark rolling out to macOS app for local tasks, automation
As previewed at I/O 2026 , Gemini for macOS now supports Gemini Spark. This is the first of two big desktop app updates coming this summer, with the other being a new voice experience.
- UBTech launches consumer humanoid robot series
UBTECH's You World U1 humanoid robots are priced from 119,800 yuan (US$18,000) to 990,000 yuan (US$146,000).
- China's UBTech launches lifelike humanoid robots for consumers
China's UBTech launches lifelike humanoid robots for consumers Nikkei Asia
- iPhone Security Fixes May Arrive Sooner as AI Speeds Up Threats
Apple is releasing some iPhone security fixes earlier as AI raises concerns about faster cyberattacks and shorter patch windows. The post iPhone Security Fixes May Arrive Sooner as AI Speeds Up Threats appeared first on TechRepublic .
- Update Your iPhone Now: Apple Just Pushed an Urgent Fix as AI-Powered Threats Increase
The tech giant has begun accelerating software patches for its iPhones and iPads.
- Apple speeds up software updates in response to AI cybersecurity concerns
Apple acknowledged that artificial intelligence is reducing the time hackers need to exploit vulnerabilities, prompting a faster update cycle
- Apple says it is releasing updates early in response to AI cybersecurity concerns
The shift marks a notable change in Apple's longstanding practice of packaging security fixes with broader software releases, an acknowledgment that AI is compressing the window attackers need to exploit known flaws.
- Why Split With Waymo In Phoenix Could 'Weigh On' Uber Stock
Uber stock could face more robotaxi fears after news that it is no longer partnering with Waymo in Phoenix, an analyst said. The post Why Split With Waymo In Phoenix Could 'Weigh On' Uber Stock appeared first on Investor's Business Daily .
- Waymo, Uber end robotaxi partnership in the Valley
Uber is no longer offering Valley customers an option to book Waymo’s robotaxis through its ride hailing app after ending a three-year pilot program intended to help both companies scale autonomous vehicle operations.
- Waymo, Uber end robotaxi partnership in Phoenix
The pilot program reached just over a dozen vehicles and helped both companies scale operations in Austin and Atlanta. Uber plans to launch another autonomous vehicle partnership in Phoenix with an undisclosed company.
- Tim Cook and EU tech chief hold ‘constructive’ virtual meeting over Siri AI standoff
Earlier today, Tim Cook reportedly held “constructive” talks with EU tech chief Henna Virkkunen as Apple and the bloc sought a way forward in their dispute over the delayed launch of Siri AI in Europe. Here are the details.
- Ford Brings Back Veteran Engineers After AI Quality Setback
Ford rehired veteran engineers after AI quality systems fell short, showing why expert oversight still matters in high-stakes automation. The post Ford Brings Back Veteran Engineers After AI Quality Setback appeared first on TechRepublic .
- Vector Institute launches open-source AI bias detector
New tool aims to flag and rework biases in writing and AI datasets. The post Vector Institute launches open-source AI bias detector first appeared on BetaKit .
- Ford’s AI failure forces it to rehire veteran engineers for car inspections
Ford’s AI failure forces it to rehire veteran engineers for car inspections Arabian Business
- Return of the ‘greybeards’: AI backfired – so Ford had to rehire humans
The US motor company found that the hundreds of AI cameras being used for design and manufacturing checks were prone to pitfalls Name: “Greybeards.” Age: There’s a clue in the name. Continue reading...
- Anthropic’s Invasion of Slack, OpenAI Cuts Inference Costs in Half, Amazon’s Higher Anthropic Costs — TITV [Video]
Anthropic’s Invasion of Slack, OpenAI Cuts Inference Costs in Half, Amazon’s Higher Anthropic Costs — TITV [Video] The Information
- Netflix used AI to put Gene Wilder's voice into a new reality show
So much for magic.
- OpenAI reportedly cut response costs for guest ChatGPT users by more than half
According to a report by The Information, OpenAI has cut inference costs for its AI models by more than half. The company applied the optimizations to ChatGPT, where the number of Nvidia GPUs needed dropped to just a few hundred at times. The article OpenAI reportedly cut response costs for guest ChatGPT users by more than half appeared first on The Decoder .
- OpenAI film ‘Artificial,’ dropped by Amazon, finds a new home with Neon
OpenAI film ‘Artificial,’ dropped by Amazon, finds a new home with Neon AP News
- Anthropic launches AI drug discovery program, joining tech giants in betting on healthcare
Anthropic will start an internal drug discovery program, part of a new push to sell artificial intelligence tools to drugmakers.
- Ride-share group BlaBlaCar taps AI for 20-country expansion
French ride-sharing and travel site BlaBlaCar said Tuesday that it would push into 20 new countries—double the current number—thanks to the scaling opportunities provided by AI.
- Autonomous driving firm Autobrains eyes robotaxi growth in Europe, CEO says
Autobrains, which is based in Tel Aviv and has an office in Munich, is developing lower-cost autonomous driving technology built around so-called agentic AI, which it says reduces reliance on expensive sensors and computing power, a key obstacle to scaling self-driving systems.
- China factory activity returns to expansion, riding AI global boom
China factory activity returns to expansion, riding AI global boom The Straits Times
- China factory activity returns to growth as AI-linked exports boom
China factory activity returns to growth as AI-linked exports boom The Straits Times
- Samsung, SK Hynix mega South Korea chips gamble tests optimism of AI cycle
ANALYSIS-Samsung, SK Hynix mega South Korea chips gamble tests optimism of AI cycle
- South Korea's mega chips gamble tests optimism of AI cycle
Samsung Electronics and SK Hynix are making one of the biggest bets yet on the artificial intelligence boom with investments worth hundreds of billions of dollars, but the planned capacity buildout is stoking fears of a painful reckoning if AI spending cools. The tech giants won praise -- and a deep bow -- from President Lee Jae Myung after throwing their weight behind the government's semiconductor push, in an apparent departure from a previously restrained approach to capacity expansion shape
- Anthropic’s Claude Science bets on workflow, not a new model, to win over scientists
Anthropic's Claude Science is a workbench that gives scientists one environment to do computational research, saving them from the need to bounce between databases, pipelines, and tools.
- Anthropic just released a brand-new Claude Science app for Mac
Anthropic just launched a brand-new desktop app called Claude Science. The new app joins the main Claude app on the Mac, which includes Claude AI, Cowork, and Code.
- Apple Creator Studio Gets New AI Features
Apple today updated its Creator Studio apps, adding new AI features to Pixelmator Pro, Final Cut Pro, Logic Pro, and more. Apple is integrating Pixelmator Pro with Final Cut Pro, Keynote, Numbers, and Pages. Final Cut Pro users can send a frame to Pixelmator Pro to create thumbnails and social graphics. In Keynote, Numbers, and Pages, users can select an image in a document and open it in Pixelmator Pro to edit, with changes saved to the original document. The three office apps support generating vector shapes using AI, and Pixelmator Pro is getting advanced image generation and a Content Hub. Users can generate AI images directly in Pixelmator Pro with natural language, and browse a curated collection of images in Content Hub. Freeform also integrates with Pixelmator Pro in iOS 27 , iPadOS 27 , and macOS Golden Gate . Final Cut Pro is getting Generate Captions, an on-device AI feature that automatically adds subtitles to videos based on audio. Subtitles can have animations and custom fonts, colors, and positions. Edit Detection is a new AI feature that analyzes rendered video and splits it back into the original clips on the timeline. Apple says editors can use the tool for edit refinements or assembling a cut-down highlight clip for social media. On the Mac, Final Cut Pro has an Auto Mask feature that isolates and refines video elements like skin, hair, sky, foliage, and clothing. Users can hover over a clip and make precise adjustments with no manual tracking. Color Match now produces more accurate and natural color matches in different lighting conditions, plus Advanced Trimming lets users fine-tune incoming and outgoing frames one-by-one. Motion gains native support for scaling vector graphics without affecting quality, and Compressor has an Immersive Metadata Viewer for the Vision Pro. Final Cut Camera is getting expanded ProRes Support, an option to disable digital zoom, and Clean HDMI Out for sending a pristine video signal to external monitors and recorders. Logic Pro's Chord ID feature has been rebuilt and it is more accurate than before. Apple says Session Players will respond and perform chord changes more quickly. Both Logic Pro and MainStage have a new granular sync mode in Alchemy to open up "new dimensions of sound design." More information on the updates can be found on Apple's website . Creator Studio Pro includes all of Apple's creative software, and it is priced at $12.99 per month or $129 per year. Up to six people can share a single membership. Tag: Apple Creator Studio This article, " Apple Creator Studio Gets New AI Features " first appeared on MacRumors.com Discuss this article in our forums
- Claude Science, an AI workbench for scientists, is now available
Claude Science, an AI workbench for scientists, is now available
- With Claude Science, Anthropic Targets Another Application
The AI vendor is not the first model provider to focus on science. It is proceeding cautiously due to challenges with using AI in the science field.
- Claude Science is Anthropic’s newest flagship product
At an event for pharmaceutical executives, biotech founders, and researchers on Tuesday, Anthropic announced Claude Science, a major new product intended to support scientific research in the same way that Claude Code supports software engineering. Like Claude Code, Claude Science can autonomously carry out meaningful work when given concise, high-level instructions, and it has access…
- Proton Lumo 2.0 Adds Image Generation, Memory, Stronger Web Search
Last summer, Swiss-based Proton launched Lumo , an AI assistant with a privacy-first approach. Today, the company has announced Lumo 2.0, a major update to the chatbot that brings three new features commensurate with its core principles of no logs, no data sharing, and zero-access encryption. Proton says Lumo 2 has been rebuilt on a new architecture that brings its biggest leap in capability to date, with Fast and Thinking modes now available. Fast of course prioritizes speed, while Thinking is optimized for more complex, multi-step reasoning. Proton says Lumo 2 responds to everyday queries up to 76 percent faster than Lumo 1.4. Beyond the new architecture, Lumo 2 also boasts multimodal capabilities such as image generation and image recognition. Users can now upload an image to analyze, create visuals from a prompt or a rough sketch, or edit existing images, all in the same conversation. On top of the new features, Lumo 2.0 has far stronger web search compared to Lumo 1.4, according to the company. There's also a Memory feature that lets Lumo learn your preferences, working style, and ongoing context, with the context window now twice as large. Lumo 2.0 also introduces Custom Lumos, described as enabling purpose-built assistants that can be tailored to specific tasks, such as a research assistant that structures answers the way you need them. Lumo is free to use at Lumo.proton.me and does not require a Proton account when accessed. However, if you have a Proton account, your chat history can be saved using the company's "zero-access" encryption across all your devices. There are also mobile apps for iPhone and Android. For power users, Lumo Plus brings unlimited chats, Projects, advanced image generation, and priority access to the fastest models. Plus costs $12.99 per month, and there's also a Lumo Professional plan for teams offering secure collaboration for $14.99 per user per month, with discounts currently available for both plans. Tag: Proton This article, " Proton Lumo 2.0 Adds Image Generation, Memory, Stronger Web Search " first appeared on MacRumors.com Discuss this article in our forums
- Anthropic launches Claude Science, an AI workspace built specifically for researchers
Anthropic released Claude Science, an AI workbench for researchers. More than 60 preconfigured skills cover fields like genomics and computational chemistry, and a verification agent automatically checks citations and calculations. The app runs locally or on HPC clusters, so sensitive data never has to leave a lab's own infrastructure. The article Anthropic launches Claude Science, an AI workspace built specifically for researchers appeared first on The Decoder .
- U.S. Eases Export Curbs on Anthropic’s Fable Model
U.S. Eases Export Curbs on Anthropic’s Fable Model The Information
- US removes curbs on Anthropic's latest Fable and Mythos AI models
US removes curbs on Anthropic's latest Fable and Mythos AI models Reuters
- AI Attention Map Reveals RNA Splicing Regulatory Regions (IMAGE)
AI Attention Map Reveals RNA Splicing Regulatory Regions (IMAGE) EurekAlert!
- The Trump Administration Is Lifting Its Export Controls on Anthropic’s Mythos and Fable AI Models
The White House is easing restrictions on Anthropic’s most advanced AI models weeks after ordering the company to suspend access for foreign nationals.
- Sonar
Understand Any Codebase, Explained for Your Role
- Claude Companion
See what Claude Code is doing — live, on your screen
- WebLegal.ai
Scan, generate and own your website's legal pages
- VCPeer
Plain-English search for VC and private market data
- Memorra
Smart planning for every celebration.
- Sakana Fugu
Sakana Fugu
- Inference chip startup Etched launches with $800M in funding
Etched Inc., a developer of artificial intelligence inference chips, launched today with $800 million in funding. The startup raised the capital over multiple rounds. The most recent investment, which closed in December, valued Etched at $5 billion. It included the participation of VentureTech Alliance, a startup fund associated with Taiwan Semiconductor Manufacturing Co. It was […] The post Inference chip startup Etched launches with $800M in funding appeared first on SiliconANGLE .
- BeatMV
Turn any song into scroll-stopping videos in minutes with AI-directed scripts and
- StemSplitterAI
Split any song into stems and custom backing tracks online.