AI News Feed

14 Nov 2025- Anthropic details methods to make Claude politically "even‑handed"—system prompts, reinforcement‑learning rewards, and an open‑source neutrality tool, reporting high even‑handedness scores versus rivals.

13 Nov 2025- OpenAI launches GPT‑5.1 with Instant and Thinking variants, expanded personality presets and user style tuning; staggered ChatGPT rollout and auto-matching to the best model.

13 Nov 2025- Yann LeCun plans to leave Meta to launch a startup building video/spatial world models, fundraising underway, amid Meta’s AI reorganization and internal tensions over technical direction.

12 Nov 2025- ElevenLabs launched the Iconic Voice Marketplace, a curated, consent‑focused platform licensing AI‑replicated voices of verified famous and historical figures (28 voices), pairing rights holders with brands for paid use.

12 Nov 2025- Google Photos brings "Help me edit" conversational AI to iPhone in US, improved face edits using private face groups, Nano Banana stylistic transforms, an "Ask" chatbot, and expanded search.

12 Nov 2025- Google’s Pixel update adds cautious AI notification summaries for long chat threads (Pixel 9+ excluding 9A), expands scam detection, VIP prioritization, AI photo “remix,” and future auto-sorting.

11 Nov 2025- OpenAI warns advanced AI is near researcher-level, predicts scientific discoveries by 2026 and breakthroughs by 2028, and urges global coordination on safety, oversight, and resilience.

11 Nov 2025- CMU and Stanford found AI agents far faster and cheaper but lower quality and prone to fabrication; human‑agent teaming with human verification preserved quality and boosted efficiency.

10 Nov 2025- Adobe MAX 2025: an AI-first creative suite—Firefly becomes a pro-grade AI studio (Image Model 5), introduces agentic assistants like Project Moonlight, plus Google Cloud and YouTube partnerships.

08 Nov 2025- Alibaba‑backed Moonshot AI released open‑source Kimi K2 Thinking, matching or beating top closed models on agentic benchmarks, chaining hundreds of tool calls; trained under $5M and priced affordably.

07 Nov 2025- Bipartisan senators urged Trump to block China’s access to advanced AI chips, models, and cloud resources, backing export controls and prioritizing allies after debates over Nvidia chip sales.

07 Nov 2025- Amazon launches Kindle Translate beta for KDP authors, AI-powered ebook translations (English↔Spanish, German→English), previews, separate pricing, labeled editions, eligible for KDP Select and Kindle Unlimited.

07 Nov 2025- Apple to license a custom 1.2T‑parameter Google Gemini for Siri, paying about $1B/year; inference on Apple’s Private Cloud, rollout possibly next spring while Apple builds its own model.

06 Nov 2025- Google’s Gemini Deep Research now accesses Gmail, Drive, and Chat to combine personal Workspace content with web search, producing synthesized research reports exportable to Docs or AI-generated podcasts.

05 Nov 2025- Microsoft launched MAI-Image-1, its first in-house text-to-image model, powering Bing Image Creator and Copilot Audio Expressions with fast, photorealistic imagery and nuanced lighting; EU rollout pending.

05 Nov 2025- Google's Project Suncatcher explores solar-powered satellite data centers hosting TPUs to run AI nearly 24/7, tackling radiation, terabit networking and debris risks, with prototypes planned by 2027.

05 Nov 2025- OpenAI signed a seven‑year, $38B AWS deal for hundreds of thousands of NVIDIA GPUs, diversifying from Microsoft and joining a $1.4T multi‑cloud infrastructure buildout.

04 Nov 2025- False reports claimed OpenAI stopped ChatGPT from giving legal or medical advice; OpenAI says behavior unchanged — unified policy only restates existing limits on tailored licensed advice.

04 Nov 2025- Ilya Sutskever's deposition in Musk's lawsuit revealed a 52‑page memo accusing Sam Altman of dishonesty, citing Mira Murati's evidence, a Brockman memo, and talks of an Anthropic merger.

03 Nov 2025- 1X Technologies’ $20K humanoid NEO is mostly teleoperated, not autonomous; launch demos exaggerated, raising privacy, data‑collection, and buyer expectation concerns.

01 Nov 2025- Meta bought nearly 1 GW of U.S. solar this week (raising 2025 commitments above 3 GW) to power and offset AI data-center demand, amid criticism over reliance on RECs/EACs.

01 Nov 2025- OpenAI unveiled a multi-year plan to build autonomous research agents—intern-level by 2026, fully automated by 2028—backed by massive compute, funding, and organizational restructuring.

31 Oct 2025- Tim Cook says Apple will integrate more third‑party AI models (ChatGPT added; Gemini, Anthropic rumored), targets AI‑upgraded Siri next year, and remains open to M&A amid record Q4 revenue.

31 Oct 2025- Man fed a $195K hospital bill into Anthropic Claude; AI flagged Medicare coding errors and extreme markups, enabling negotiation that cut charges to $33K—about $162K saved.

30 Oct 2025- Zuckerberg says AI will add vast new content to Meta’s recommendation engines; Meta embeds AI, tests AI social apps (Vibes: 20B images) while reporting $51.24B revenue and $15.93B tax charge.

30 Oct 2025- OpenAI became a public benefit corporation; new OpenAI Foundation holds ~$130B (allocating $25B to health/AI resilience), Microsoft stake ~27% (~$135B), AGI verification panel, 2032 tech rights.

29 Oct 2025- Bipartisan GUARD Act would ban under‑18s from AI chatbots, mandate age verification (IDs/face scans), require 30‑minute disclosures they're not human, forbid sexual/suicide‑promoting content, and impose penalties.

28 Oct 2025- xAI launched Grokipedia, a Wikipedia-like encyclopedia that reuses Wikipedia content, restricts public edits, sometimes alters facts (e.g., climate change), and raises copyright and accuracy concerns.

28 Oct 2025- About 600 ex‑Meta hires — roughly 20% of staff — are reshaping OpenAI’s culture, sparking tensions over growth strategies, ad‑adjacent ideas, leadership departures, and "becoming Meta."

27 Oct 2025- UK analyst Julien Garran warns today's AI boom is the largest, riskiest bubble—vast unprofitable valuations, funding fragility, potential economic drag; he doubts LLM-driven commercial success.

25 Oct 2025- Sam Altman’s Merge Labs hired Caltech engineer Mikhail Shapiro to lead noninvasive BCI work using ultrasound, magnetic and gene‑based techniques; company seeks large funding and favors read‑only interfaces.

25 Oct 2025- Microsoft relaunched agentic Copilot Mode in Edge—an Atlas‑like browsing assistant that reasons over tabs and acts for users—escalating competition and raising privacy and monetization concerns.

24 Oct 2025- EA partners with Stability AI to co-develop AI tools for game content creation — from PBR texture generation to 3D previsualization — augmenting artists while keeping humans central.

24 Oct 2025- Meta, Amazon and Alibaba unveiled AR/AI glasses: Meta's glitchy demo; Amazon's logistics-focused hands-free scanner/controller; Alibaba's consumer Quark glasses with Qwen LLM priced $660, betting on wearables.

24 Oct 2025- Microsoft launches Edge's "Copilot Mode" — tab-based AI chat integrating search and navigation, adding preview Copilot Actions and Journeys; limited rollout with mixed, error-prone agentic features.

22 Oct 2025- YouTube rolls out AI "Likeness Detection" to Partner Program creators to flag videos using their face, letting verified creators review and request removal of suspected deepfakes.

22 Oct 2025- Anthropic launched Claude Code: a sandboxed coding agent isolating filesystem and network access, cutting permission prompts 84%, proxying git credentials, running isolated cloud sessions, and open‑sourcing the sandbox.

21 Oct 2025- Friend’s $129 chatbot pendant sparked NYC backlash after a $1M subway ad campaign; protests and graffiti escalated, founder Avi Schiffmann confronted critics and pledged not to sell to Big Tech.

21 Oct 2025- AI-generated Bryan Cranston videos on OpenAI’s Sora 2 led actors, agencies and SAG‑AFTRA to demand protections; OpenAI pledged stronger guardrails and opt‑in likeness controls, but gave no specifics.

20 Oct 2025- Andrej Karpathy says no AGI until 2035, criticizes LLMs' cognitive limits (RL rewards lucky guesses, coding bloat, memorization), warns of reliability costs, predicts a decade of augmented intelligence.

18 Oct 2025- Meta is rolling out an opt‑in Facebook feature in US/Canada that uploads camera‑roll photos to the cloud to suggest edits — saying uploads won't train AI unless edited or shared.

18 Oct 2025- Anthropic launches "Skills" for Claude: folder-based, dynamically loaded bundles of workflows, scripts, and assets, plus an interactive creator, enabling enterprise automation like on‑brand PowerPoint generation.

17 Oct 2025- Google launches Veo 3.1 — editing-focused video generation supporting three reference images, start/end-frame transitions with audio, and scene extension to ~1-minute; available in Flow, Vertex AI, and Gemini.

17 Oct 2025- Anthropic forecasts ~$9B ARR, targets $25B in 2026, launches Claude Haiku 4.5 (faster, cheaper); Google unveils Veo 3.1 video-generation with character consistency and advanced editing.

16 Oct 2025- Google updates Flow with Veo 3.1: improved motion, lighting and shadow controls, new audio generation (Ingredients/Frames/Scene Extension), object-removal, available as paid preview via Gemini API/app.

16 Oct 2025- Stanford’s AI Evo, trained on ~2M phage genomes, generated synthesized bacteriophages—16 of 302 proved functional against E. coli, promising faster phage therapies while raising safety concerns.

15 Oct 2025- OpenAI will permit erotica in ChatGPT for verified adults starting December, adding age‑verification and controls; Altman cites relaxed restrictions and formed a well‑being expert council.

14 Oct 2025- California's SB 243 requires companion chatbots to clearly disclose they're AI and, starting next year, submit annual suicide‑prevention reports to the state Office of Suicide Prevention.

14 Oct 2025- Andrew Tulloch, co‑founder of Thinking Machines Lab, left to rejoin Meta amid reported multibillion-dollar recruitment, underscoring intense competition for AI talent and Meta’s AI reorganization.

13 Oct 2025- David Pierce says OpenAI’s ChatGPT is shifting into an “operating system” hosting apps, turning the assistant into an everything app while stirring cultural debates about AI art and games.