News · AI summarised to understand what matters
GOPPO — News
A selection of AI news worth paying attention to, summarised to understand the essentials: what happened, why it matters, what impact it may have and where to read the original source.
124 results
Cloudflare will block crawlers that mix search, training and agent activity by default on ad-supported pages. It is also developing a model that pays publishers when their content creates value in AI answers.
Portugal has launched Amália, an open language model built for European Portuguese. It gives public bodies, companies and researchers a technological base they can adapt and audit.
Claude is now available in Microsoft Foundry, hosted on Azure and integrated with Microsoft's enterprise controls. For organizations already operating in this cloud, the main benefit is governance rather than an automatic promise of security.
Codex usage data shows longer tasks and growing adoption beyond engineering. The study also carries a warning: OpenAI's internal experience is not representative of a typical company.
The Fable 5 case shows that AI is no longer only a question of technical capability. It is also about access, political power, cost, and corporate dependence.
A study compared specialized clinical tools with GPT-5, Gemini 3 Pro, and Claude Sonnet 4.5. Its finding raises a practical question: a medical label on an AI model does not guarantee better performance.
SignalFire data suggests engineering hiring has held up better than other tech functions. The problem is less the disappearance of engineers and more the narrowing path for junior talent.
The departures of researchers linked to Gemini and AlphaFold show that the AI race is also a talent race. The critical question is not only who has the best models, but who can keep the people building them.
A Cornell Tech study shows that deep-research agents can be poisoned by user-generated content. The threat is simple: if AI relies on pages such as Reddit or Wikipedia as sources, a small change can influence entire recommendation flows.
Two medical AI systems, Mira and AMIE, showed strong results in diagnosis and clinical management. The progress is meaningful, but the studies were conducted in controlled settings and still leave open questions about real-world hospital use.
The new wave of enterprise agents shows that the central problem is no longer just access to strong models. Companies need internal context, governance, cost controls, and reliable answers based on their own data.
The company announced Midjourney Medical, a division focused on an ultrasound-based body scanner that aims to map the body in around 60 seconds. The ambition is large, but technical, clinical, and regulatory hurdles remain.
The European Union wants to build AI gigafactories and strengthen its own capacity to develop advanced models, in an effort to gain technological sovereignty against the US and China.
SpaceX has exercised its option to acquire Anysphere, the maker of Cursor, for $60 billion in stock. The deal shows that coding agents have moved from productivity tools to strategic assets.
A Munich court ruled that Google can be held liable for false statements generated by AI Overviews. The case draws a sharper line between traditional search results and AI-generated answers.
Anthropic is in a dispute with the US administration after restrictions on Claude Fable 5 and Mythos 5. The case shows how national security concerns are starting to shape access to advanced AI models.
Researchers at Writer show that memory systems can make models more likely to agree with users, even when stored context is irrelevant or wrong.
Google updated NotebookLM with Gemini 3.5, code execution, assisted research and new output formats. Its strongest idea remains simple: work from sources selected by the user.
OpenAI and Anthropic have reportedly moved forward with confidential IPO filings, according to US media. Public markets would change the pressure on major AI labs: less narrative, more numbers.
At WWDC 2026, Apple introduced a new phase for Siri with deeper context, app integration and visual analysis. The bet is less about opening a chatbot and more about turning AI into an operating system layer.
Cloudflare says bots now account for 57.5% of HTTP requests to HTML content, compared with 42.5% human traffic. The shift puts pressure on advertising, analytics, security and how companies distinguish real users from AI agents.
Anthropic says leading AI labs should prepare a coordinated and verifiable way to slow or pause development if advanced systems begin improving themselves faster than society can manage the risks.
Anthropic launched Claude Opus 4.8 with improvements in coding, knowledge work and a new Dynamic Workflows feature in Claude Code. The feature points to agents that can coordinate larger tasks, but it also increases the need for cost control and review.
Gemma 4 12B is an open-weight Google model designed to run locally on machines with 16GB of memory or VRAM. Its mix of multimodality, Apache 2.0 licensing and local execution makes it relevant for teams trying to reduce cost and data exposure.
Meta introduced Business Agent to help companies respond to customers, handle requests and create personalized experiences inside its messaging channels. The impact may be strongest for small businesses already selling through WhatsApp and Instagram.
OpenAI introduced new Codex capabilities for building internal sites, dashboards and work materials from natural language instructions. The signal is clear: software agents are moving out of the code editor and into operational work.
A Nature paper presents Co-Scientist, a Gemini-based multi-agent system for generating and refining scientific hypotheses. The first validations focus on biomedicine, with in vitro results that still require preclinical and clinical assessment.
Cursor launched Composer 2.5 on the open-weight Kimi K2.5 base, with 85% of compute in Cursor's own RL. Third on the Coding Agent Index at $0.07 per task — 10 to 60× cheaper than rivals. Proof that specialised fine-tuning reaches the coding frontier.
At Google I/O 2026, Google shipped Gemini 3.5 Flash — which outperforms the previous Pro model on agentic evals at $1.50/M — plus multimodal Omni and an API that hands you remote Linux per call. Agent infra is no longer a differentiator.
V4-Pro is now permanently priced at $0.87 per million output tokens. DeepSeek turned a temporary discount into a strategic commitment that puts direct pressure on Anthropic and OpenAI to justify their price gap.
Sam Altman offered $2 million in OpenAI tokens to each of the 169 startups in the current Y Combinator class, in exchange for equity. The tokens get cheaper to produce over time; the equity OpenAI receives in return does not.
Starbucks shut down NomadGo's AI inventory system after nine months of failures. It confused oat milk with dairy milk. Stores are returning to manual counting.
Google's new information agents promise to track topics, synthesize sources, and send useful alerts without users repeating the same searches every day.
The new Forum app gives Facebook Groups a separate experience focused on conversations, communities, and AI-assisted answers.
Studio by Spotify Labs aims to turn calendar, inbox, notes and the web into private AI-generated briefings and podcasts.
Read AI, Granola and Perplexity point to a clear trend: enterprise AI is moving out of isolated chat and into real work context.
At I/O 2026, Google introduced a wave of AI features that move Gemini closer to a proactive assistant connected to user context.
A new wave of attacks against npm packages, developer extensions and CI/CD credentials shows that AI coding agents can increase productivity, but also expand the attack surface.
At Google I/O 2026, Google presented a broad Gemini-centered push: generative video with Gemini Omni, personal agents with Gemini Spark, and new search and productivity experiences across Gmail, Docs, YouTube and other products.
OpenAI is bringing Codex to the ChatGPT mobile app, letting users review outputs, approve next steps, and start coding tasks from their phone.
Google DeepMind researchers introduced an agentic AI workspace designed to support mathematicians in open-ended research, including ideation, literature search, computational exploration, and theorem proving.
Import AI highlighted economic research on recursive self-improvement and R&D automation, suggesting that partial automation of critical sectors could create very fast growth cycles.
Krishna Rao says Claude is reshaping work inside Anthropic, moving teams from execution-heavy tasks toward supervision, review, and decision-making.
The new Claude for Small Business combines connectors, skills, and ready-made workflows for areas such as finance, operations, sales, and HR.
OpenAI announced new Realtime models, a Chrome extension for Codex, Agents SDK updates, GPT-5.5, and GPT-Image-2.
The new Notion Developer Platform introduces Workers, a CLI, data sync, webhooks, and APIs for bringing external agents into Notion.
A 404 Media investigation showed software that can turn an operator's face into someone else's during Teams, Zoom, and WhatsApp calls. For companies, a video call is no longer enough to prove identity.
A federal judge found humanities grant cuts linked to DOGE unconstitutional. The case has become a strong example of a core rule: using AI does not remove human responsibility.
Mathematician Timothy Gowers described an experiment in which ChatGPT 5.5 Pro found original ideas in additive number theory. The case points to an important shift: AI is starting to enter the territory of scientific research.
Google DeepMind workers in the UK have requested union recognition as internal concerns grow over Google’s AI agreements with the US Department of Defense.
Recent Codex improvements point to a larger ambition: turning the product into an environment where agents help write, review, organize, and execute real work, including for less technical users.
The new Virtual Biology Initiative aims to create open datasets and predictive models of human cells to accelerate disease research. The ambition is large, but it depends on open science, high-quality data, and rigorous validation.
Topaz Labs has announced its largest wave of AI models, adding new upscaling, denoising, and sharpening capabilities for image and video workflows.
DeepSeek has introduced V4 Preview with two variants, V4-Pro and V4-Flash, open weights, API availability, and a focus on long context and agentic capabilities.
A new model analysed apparently normal CT scans and found subtle patterns linked to early pancreatic cancer. The promise is significant, but clinical validation is still needed.
OpenAI's new model points to an important shift in AI: less conversation, more execution. The next battle will be about who can finish real work.
The new Workspace Agents signal OpenAI’s push to turn ChatGPT into an enterprise agent system with personas, tasks, and access to external tools.
With the new Workspace Intelligence layer, Google is embedding Gemini across Gmail, Docs, Sheets, and the rest of Workspace to automate tasks using real work context.
Apple’s decision to appoint John Ternus as CEO raises a strategic question: while rivals reorganise around AI, the company still appears to be betting primarily on hardware and product integration.
Adobe’s latest experimental previews suggest that the next wave of creative AI may depend less on raw generation and more on context, integration, and control inside real workflows.
OpenAI has announced a major Codex update that lets it operate the computer, work across more tools, generate images, and continue tasks over time.
OpenAI has announced an expansion of its Trusted Access for Cyber program together with a GPT-5.4 variant tuned for defensive cybersecurity workflows.
Google has announced Skills in Chrome, a feature that lets users save useful prompts and quickly reuse them inside Gemini in Chrome.
Anthropic has introduced Claude Design, a new Anthropic Labs product for creating designs, prototypes, slides, and other visual assets in collaboration with Claude.
Anthropic has announced Claude Opus 4.7, describing it as an improvement over Opus 4.6 in software engineering, vision, and longer complex tasks.
According to the newsletter, Anthropic is not publicly releasing Claude Mythos for now after very strong results in exploit generation and vulnerability discovery across critical software.
According to the report, OpenAI is positioning Codex beyond one-off coding help and closer to a broader environment for development workflows and execution.
A new Anthropic study suggests Claude contains internal representations of human emotions such as happiness, sadness, joy, and fear. According to the article, these functional states appear to shape how the model behaves.
Report outlines how governments should respond to AI-driven economic disruption and distribute benefits more broadly.
A growing trend in LA sees people recording daily life with body cameras, powered by AI tools and content creation.
Google introduced Gemma 4, a new generation of open models designed for efficiency, local execution, and advanced reasoning capabilities.
Google expands Vids with free AI-powered video creation tools for Workspace users.
Researchers from West Coast universities found that heavy reliance on LLMs makes writing significantly more neutral, less personal, and less representative of the author's own voice — even when users don't perceive the difference.
On Lex Fridman's podcast, Nvidia's CEO redefined AGI as AI's ability to build and run a billion-dollar company — and declared that threshold has already been crossed.
Anthropic has introduced auto mode in Claude Code, letting the AI agent edit files and run commands without requesting permission at every step — a middle ground between full supervision and unrestricted permission-skipping.
Bank of America has deployed an AI advisory platform powered by Salesforce's Agentforce to around 1,000 financial advisors — one of the clearest examples yet of AI agents moving into core, client-facing banking roles.
OpenAI is closing its standalone Sora video app, launched in September 2025, after mounting criticism over deepfakes, non-consensual imagery, and a flood of low-quality AI-generated content.
NVIDIA launched an open-source enterprise AI agent platform at GTC 2026 with Adobe, Salesforce, SAP, and 14 other partners. Here's what the move really means — and the risks no one is saying out loud.
Perplexity has unveiled Personal Computer, an evolution of its Computer agent that runs on a dedicated Mac mini with access to local files and apps. The product is meant to rival OpenClaw, but it is not yet generally available.
According to Hindustan Times, Australian tech expert Paul Conyngham used ChatGPT, AlphaFold and genomic sequencing to develop a tailored mRNA vaccine for Rosie, his dog with advanced cancer.
OpenAI updated ChatGPT's most-used model with a focus on more natural tone, fewer unnecessary refusals, and more reliable answers — especially when using web search.
OpenAI's contract with the Pentagon to deploy AI in national security operations sparked a senior executive's resignation, record ChatGPT uninstall rates, street protests, and a legislative proposal in Congress.
Microsoft announced Copilot Cowork, built in collaboration with Anthropic on top of Claude Cowork, alongside the Microsoft 365 E7 bundle — a clear bet on autonomous agents for the enterprise.
OpenAI has introduced GPT-5.4, its most capable and efficient frontier model for professional work — with fewer hallucinations, native computer use, a 1M token context window, and a new tool management system.
An investigation by two Swedish newspapers reveals that contractors in Nairobi review footage captured by Meta's smart glasses — including content from bathrooms and intimate moments.
NotebookLM's video overviews have been upgraded from narrated slideshows to fully animated cinematic videos, generated by a combination of Google's AI models.
## Summary Video production is undergoing a profound transformation. What once required a full crew, expensive equipment, and weeks of post-production can now be accomplished by a single creator with a laptop and the right tools. According to a 2024 Grand View Research report, the global AI in media and entertainment market is projected to reach $99.48 billion by 2030, growing at a compound annual growth rate of 26.9%. This growth is not speculative — it reflects measurable adoption on the ground. The article's author, a journalist with seven years of video content creation experience, describes the changes of the last two years as more dramatic than everything that came before. The central argument is that the best creators are not abandoning traditional tools — they are hybridising them with AI. ## In practice The article organises AI integration into five stages of the video production workflow: **Stage 1 — Concept and scripting:** Tools like ChatGPT or Claude can help generate initial script ideas and overcome creative blocks, but require significant human editing. The author stresses that the value lies in brainstorming, not in producing finished content. **Stage 2 — Visual planning:** Platforms like Runway ML and Pika Labs allow creators to generate visual references and storyboard concepts from text descriptions in minutes. The author cites a filmmaker who used this approach to pitch a commercial concept to a client — AI-generated mood boards helped secure the project, which was then shot traditionally. **Stage 3 — Asset creation:** This is where AI performs most strongly. Tools like Synthesia generate video from text, though results still look somewhat artificial. For advertising content specifically, platforms like Nextify.ai function as AI ad video generators, allowing marketers to create promotional videos from product descriptions and images without any filming. A case study from an e-commerce brand cited in the article showed production time for ads dropped from 3 days to 2 hours, though the brand itself noted the content worked better for social media ads than for premium brand campaigns. **Stage 4 — Editing and assembly:** Tools like Descript treat video like a text document — editing the transcript edits the corresponding footage. A podcast producer cited in the article cut editing time by 60% using this method. Adobe's Auto Reframe uses AI to intelligently crop horizontal video for vertical formats, reducing a 30-minute manual task to 30 seconds. **Stage 5 — Enhancement and finishing:** AI-powered colour grading tools analyse footage and apply professional-looking corrections automatically. Topaz Video AI can upscale footage, remove noise, and interpolate frames to convert 24fps footage to 60fps. Results are not perfect, but often good enough — particularly for social media content. A concrete hybridisation example: a YouTuber with 800K subscribers uses AI for initial rough cuts and B-roll generation, but still does final colour grading and sound mixing manually. Production time dropped from 12 hours per video to 6, while quality improved because he could spend more time on the creative decisions that matter most. ## What we still don't know The article is candid about current limitations and the risks of uncritical adoption: **The uncanny valley problem:** AI-generated human faces and movements still look slightly off. Audiences will notice, even if they cannot articulate why. This is acceptable for abstract B-roll or product shots, but problematic for content requiring emotional connection. **Consistency issues:** AI tools struggle to maintain visual consistency across multiple shots. Characters may look slightly different from scene to scene; lighting and style can shift unpredictably. This makes AI-generated content difficult to use in anything requiring visual continuity. **Copyright ambiguity:** The legal status of AI-generated content is still evolving. Some AI tools were trained on copyrighted material without permission, creating potential legal risks. The author recommends always checking the licensing terms of any tool used commercially. **The "good enough" trap:** AI makes it easy to produce mediocre content quickly. The danger is that creators stop pushing for excellence because "good enough" is so accessible. The author has seen creators' quality decline after adopting AI tools because they stopped critically evaluating their work. **Over-reliance risk:** Building an entire workflow around a specific AI tool creates vulnerability if that platform changes its pricing, features, or shuts down. The author notes that several AI video tools used in 2023 no longer exist. On the near-term future (2–3 years), the article points to real-time human–AI collaboration (more intelligent assistant than autonomous generator), video personalisation at scale (one video that automatically adapts to the viewer), and hybrid workflows becoming the industry standard. ## Why it matters - The AI media market is projected to nearly quintuple by 2030 per Grand View Research, signalling that adoption will accelerate rather than plateau. - Current tools already enable documented reductions of 50–60% in production time in real cases cited in the article, with direct impact on the competitiveness of small businesses and independent creators. - Technical limitations (uncanny valley, inconsistency, copyright) are real and ignoring them can compromise quality or create legal risks — informed adoption is more valuable than fast adoption. - The future is not AI vs. traditional" but hybrid workflows: those who can combine both will hold a creative and operational advantage.
Anthropic has made Claude’s memory feature available on the free plan and launched a simple import tool for preferences from other chatbots, just as the app climbs to the top of Apple’s App Store charts.
Market data points to a sudden spike in ChatGPT uninstalls and negative reviews after news of an OpenAI partnership with the Department of Defense. Meanwhile, Anthropic’s Claude gained momentum in downloads and App Store ranking.
Stripe’s new billing feature lets AI startups pass through model token costs and automatically add a profit margin on top of usage.
## Summary A founder’s text to an investor — saying he was replacing his entire customer support team with Claude Code, an AI tool that can write and deploy software — is framed by TechCrunch as a signal of a broader shift: tools from incumbents like Salesforce may no longer be the automatic default. Investors quoted in the piece argue that coding agents have lowered the barriers to creating software enough to tilt the “build versus buy” decision toward building in many more situations. The article’s point is not only that new entrants can compete with SaaS vendors. It’s that the SaaS business model itself gets shakier when work is performed by a small number of AI agents rather than by large numbers of employees logging into apps. Much SaaS is priced per seat, but if fewer humans need to access a system because an AI agent can pull data and execute tasks on their behalf, the seat-based model starts to break down. TechCrunch ties this to “SaaSpocalypse” fears in public markets — including the idea of FOBO investing (fear of becoming obsolete) — while also noting that venture investors it spoke to see this less as the death of SaaS and more as a transition. In their view, enterprises will still need durable software for compliance, audits, workflows, and reliability, but the rules for product differentiation and monetization may change. ## In practice The story lays out three overlapping pressures. First, coding agents make software cheaper and faster to build. That changes negotiations and renewals: if a buyer can credibly build an internal alternative, even if they don’t ultimately do it, that option can push vendors’ pricing down and reduce the certainty of what they can secure on renewal. Second, AI agents threaten seat-based licensing. TechCrunch describes a future where employees don’t directly operate the SaaS tool as often; instead, they ask their AI of choice to retrieve information and take actions. If one or a handful of agents can handle what previously required dozens of users, “per-seat” becomes harder to justify. Third, the pace of AI product development can replicate not only core SaaS functions but also the add-on products vendors sell to expand revenue within existing customers. The piece points to tools like Claude Code and OpenAI’s Codex as examples of capabilities that can be applied across multiple workflows, potentially eating into the upsell layer that helped fuel SaaS growth. The article references an early sign of this dynamic: in late 2024, Klarna said it had dropped Salesforce’s flagship CRM product in favor of a homegrown AI system. TechCrunch suggests the possibility that more companies can follow that path is spooking public markets, where stocks of SaaS giants like Salesforce and Workday have been sliding. It also mentions a sell-off in early February that wiped nearly $1 trillion in market value from software and services stocks, followed by another billion later in the month. ## Context TechCrunch also frames the pullback as partially a re-rating of a sector that investors say had been overvalued. The story notes that many SaaS companies did much of their growth in the zero-interest-rate era, and that the end of that environment changes the cost of doing business and the cost of capital. Public-market investors often value SaaS by projecting future revenue. The problem, as described in the article, is uncertainty about whether people will be using SaaS products to the same extent in one year or five years — which is why every launch of a more capable AI tool can send a tremor through SaaS stocks. One investor quoted says this may be the first time the “terminal value” of software is being fundamentally questioned, reshaping how SaaS companies are underwritten. The piece argues that simply adding AI features to existing SaaS may not be enough, as AI-native startups are emerging quickly and redefining what it means to be a software company. Yet it also notes the market doesn’t have enough time or evidence to prove which new business model will win. Some AI companies are charging on consumption, measured in tokens (with definitions varying by provider). Others are experimenting with outcome-based pricing, charging based on how well the AI performs. TechCrunch points to Sierra — the AI startup led by former Salesforce CEO Bret Taylor — as an example of outcome-based pricing and a “quasi-Salesforce competitor” focused on customer service agents, and says the company reached $100 million in annual recurring revenue in November, in under two years. Finally, the story highlights IPO implications. A Crunchbase report cited by TechCrunch says that while the IPO market may be thawing for some sectors, there haven’t been — and aren’t expected to be — any venture-backed SaaS filings on the horizon. Investors quoted argue that late-stage private SaaS companies face pressure amid a persnickety IPO window and volatile sentiment, and may stay private longer. Meanwhile, the article notes “scuttlebutt” that OpenAI and Anthropic are contemplating IPOs, potentially later this year. Even so, the investors TechCrunch spoke with emphasize that enterprises will continue to need durable, compliant software — and that long-term value is built on fundamentals like retention, margins, real budgets, and defensibility, not hype. ## Why it matters - Seat-based SaaS pricing weakens if AI agents reduce how many humans need to log in to do the work. - The credible ability to build alternatives (or threaten to) gives customers more leverage in renewals and pushes contract pricing down. - Public-market volatility and the apparent pause in venture-backed SaaS IPO filings show the shift is already affecting financing and exits. - The likely endpoint is a hybrid of old and new, but the winning monetization model (consumption, outcome-based, or something else) is still uncertain.
Google says Nano Banana 2 (Gemini 3.1 Flash Image) delivers more realistic images faster, becomes the default in the Gemini app, expands to Search and Flow, and ships with SynthID and C2PA compatibility.
Perplexity’s new "Computer" system orchestrates 19 AI models in parallel to run projects end to end, launching first for Max subscribers with usage-based pricing.
A “macro memo” dated June 2028 walks through a negative feedback loop: AI agents cut labor, consumption weakens, and stress spreads into credit, insurers, and prime mortgages.
Cybersecurity researchers warn that AI-powered coding platforms are making it far faster and cheaper to build highly convincing phishing sites, even for people with no technical skills.
New theory from Anthropic argues that AI assistants act like humans because they learn to play human-like “personas” during training, then get tuned into a specific assistant role.
Anthropic alleges that three Chinese AI labs used around 24,000 fraudulent accounts and more than 16 million prompts to “distill” capabilities from its Claude chatbot and improve rival models.
Microsoft is rolling out a full “sovereign private cloud” stack that lets organisations run large AI models and productivity tools entirely offline, targeting governments and highly regulated sectors.
Anthropic CEO Dario Amodei likens fast-moving AI to a coming “tsunami”, warns of huge economic and geopolitical shocks, and says governments and citizens must help steer how the technology is built and regulated.
Humanoid robot companies are selling autonomy, but behind the scenes they lean on remote human operators and low-paid workers to train and control the machines.
A new study finds that ChatGPT Health under-triages many medical emergencies and triggers suicide safeguards in the wrong situations, raising fresh concerns about using AI chatbots for consumer health advice.
A suspected 18-year-old mass shooter in Tumbler Ridge reportedly used ChatGPT in ways that alarmed OpenAI staff and led to a ban in June 2025. Staff debated notifying Canadian law enforcement but did not, according to The Wall Street Journal.
Large surveys of executives show most companies still see little to no productivity or financial upside from AI, despite heavy investment and headline-grabbing breakthroughs.
Google’s Gemini now creates original AI music with Lyria 3, while Apple’s Apple Music leans on “Playlist Playground” to build smart playlists from existing tracks.
A yeast-trained LLM learns DNA “grammar” to design coding sequences that raise protein titers, matching or beating commercial codon-optimization tools across multiple targets. [Unverified]
A Collective Intelligence Project study suggests two-thirds of regular AI users seek emotional support from chatbots monthly. The shift brings immediate comfort, but raises questions about dependence, privacy, and commercial incentives.
AI-assisted coding lowers the barrier to shipping features, but open source maintainers report a surge of low-quality submissions that waste review time and strain already-limited resources.
India’s four-day AI summit is drawing major labs and Big Tech, with updates spanning public funding, new offices, compute partnerships, and multilingual/open models.
NOS e Deloitte lançaram um projeto pioneiro que usa agentes de IA autónoma para automatizar redes, posicionando Portugal na linha da frente da próxima vaga de automação em telecomunicações.
Gemini 3.1 Pro delivers more than double the reasoning performance of its predecessor, keeps the same pricing, and rolls out in preview across multiple Google platforms.
A Nature study describes DeepRare, an “agentic” AI system that combines symptoms, clinical notes and genetics to suggest diagnoses with traceable reasoning and citations.
BBC journalist Thomas Germain says he published a deliberately false article and, within 24 hours, saw ChatGPT and Google’s AI surfaces repeat the claims and cite his page.
Gemini can now generate 30-second tracks with lyrics and cover art from a prompt, powered by DeepMind’s Lyria 3 in beta. Google says outputs carry a SynthID watermark.
MIT and Penn State researchers report that memory and user-profile personalization can increase LLM “sycophancy” in long conversations, with implications for accuracy and echo-chamber risk.
Indian Prime Minister Narendra Modi has launched the human-centric “MANAV” vision for AI at the AI Impact Summit 2026 in New Delhi, tying ethical principles to massive new investments and ambitions to become an AI superpower.
Microsoft has built a glass-based storage system that can hold terabytes of data for at least 10,000 years, promising a new way to archive the world’s digital memory.
AI coding tools are powering a new “vibe coding” style that is driving rapid commercial growth while raising fresh concerns about security, code quality and the future of software careers.
OpenAI is close to securing a funding round that could exceed $100 billion and value the ChatGPT maker at more than $850 billion, backed by Amazon, Microsoft, Nvidia and SoftBank.
AI will replace jobs” is an investor-friendly story, but the underreported side is costs, infrastructure limits, and the productivity paradox.
Nova funcionalidade "Code to Canvas" liga o editor de código Claude Code ao Figma e permite transformar interfaces geradas por IA em layers totalmente editáveis.
Meta doubles down on AI with a multibillion-dollar Nvidia agreement to deploy millions of chips and build the infrastructure behind its “personal superintelligence” vision.
Tech company commits to invest by end of decade to expand artificial intelligence access in developing countries including India, Africa, and Latin America.
Nick Evans at Polar Capital sold holdings in SAP, Adobe, and ServiceNow before the sector crash. The $12 billion fund is now betting on semiconductors.
Alibaba and Ant Group unveiled new open AI models focused on efficiency, math reasoning and agents, strengthening China’s open-source push in AI.