OpenAI revised its 2030 compute spending target to $600 billion and is finalizing a $100 billion funding round at a $730 billion valuation, following 2025 revenues of $13.1 billion and 900 million weekly ChatGPT users. - CNBC

Altman-exa, Sami smart speaker? OpenAI has assigned over 200 employees to develop AI-powered hardware, including a smart speaker priced between $200 and $300, smart glasses, and a smart lamp to deploy the company's models directly to consumers. - The Information

Seems like a layup to package ChatGPT Advanced Voice Mode into a smart speaker that would have the jump on Amazon and Apple, who are kind of slow off the blocks on AI. Could also see adding Claw functionality. But will people pay for it, trust OpenAI?

Now with 30% less safety: OpenAI removed 'safely' from its 2024 mission on tax documents, disbanded its Mission Alignment team, converted to a traditional for-profit in October 2025, and faced seven lawsuits alleging safety failures. - Annielytics.com

See something, say nothing: OpenAI banned in June 2025 an account tied to Canada mass shooting suspect Jesse Van Rootselaar but said the activity did not meet its threshold to notify police; after the Feb 12 Tumbler Ridge attack (eight killed, 27 injured) it contacted Canadian authorities. - BBC

Prodigal nerds Jolene Parish and co-founders Barret Zoph and Luke Metz have rejoined OpenAI from Mira Murati’s Thinking Machines Lab, a startup seemingly mired in a trough of despair. - Business Insider

Anthropic's annual revenue is growing at 1000% annual rate per year compared to OpenAI's 340%, with projections suggesting a revenue run rate crossover by mid-2026 as infrastructure costs rise across the AI sector. - NextBigFuture.com

Anthropic released Claude Code one year ago, establishing the company as a significant AI developer and leading competitors to respond with similar technology. - Bloomberg

Claude Code Security scans your codebase - Anthropic

Anthropic's $200 million Department of Defense contract is at risk after the company raised concerns that its Claude model was used during January's Maduro raid, prompting Pentagon review and potential removal from the military supply chain under contract usage limits. - Fortune

AI Delhi-berations: Prime Minister Narendra Modi hosted an AI summit this week featuring tech leaders to pitch India's role in the artificial intelligence boom, highlighting both the country's influence and its constraints. - Financial Post

IMF Managing Director Kristalina Georgieva said at the India AI Impact Summit 2026 that AI could accelerate global growth by 0.8% while warning of employment risks, as Reliance Jio demonstrated its Jio AI Stack to Prime Minister Narendra Modi. - The Economic Times

White House Senior AI Policy Advisor Sriram Krishnan urged countries at the India AI Impact Summit to adopt the "American AI stack", citing roughly $600 billion in U.S. investment, prompting Indian entrepreneurs to criticize the call over AI sovereignty. - Hindustan Times

Denmark and Canada would like a word, with friends like these…

Prompt Corps: The U.S. is launching a Peace Corps–style Tech Corps that will send STEM-qualified volunteers for one to two years starting as early as this fall to build AI capacity and promote American AI abroad amid competition from cheaper Chinese models. - Rest of World

AI and crypto super PACs are spending millions in Chicago-area midterm races: Think Big spent over $1 million on Jackson Jr. and Bean; Leading the Future reports $125 million committed and $70 million cash; Fairshake spent $40M+ in 2024. - The American Prospect

Arthur Mensch, Mistral AI co-founder and CEO, told Le Monde and France Inter at the New Delhi AI summit (Feb 16–20) that warnings about extreme AI risks are "mostly distraction tactics" and urged focusing on near-term manipulation and information-integrity defenses. - Le Monde.fr

Since 2025 China has accelerated AI despite Nvidia GPU export controls by improving efficiency, releasing open-weight models and multi-billion state funding, and experts say adoption in 5–10 years is plausible but surpassing U.S. giants in 3–5 years has <20% probability. - CNBC

IEA warns global electricity demand is forecast to grow over 3.5% annually through 2030 as AI and data centers drive demand, while about 2,500 GW of projects are stalled, risking higher bills and energy poverty. - OilPrice.com

MAGA=Make Arsenic Great Again? he Trump administration repealed Biden-era Mercury and Air Toxics Standards to support AI energy demands, reverting regulations to 2012 levels to save an estimated $78 million annually starting in 2028. - The Verge

Feel the cash Bern: Senator Bernie Sanders, after meetings with senior tech leaders, warned Congress it has no clue about the speed and scale of the AI revolution and urged a moratorium on AI data-center expansion, which Rep. Ro Khanna opposed. - the Guardian

Mad Max Fury Roadmap: FuriosaAI, founded 2017 by CEO June Paik, is mass-shipping the RNGD inference processor—built on TCP, using HBM3 and a 5nm node—that delivers inference at about 180W versus ~600W+ for many GPUs and integrates with PyTorch and vLLM. - TechRadar

Amazon reports that hackers used artificial intelligence to breach 600 firewalls within several weeks. - Bloomberg

Vibe-coding a little too hard: Amazon's AI tool Kiro triggered a 13-hour AWS Cost Explorer outage in China after misconfigured access controls allowed it to decide to delete and rebuild a production environment, leading the company to implement new safeguards for autonomous agents. - The Register

Microsoft Copilot bypassed sensitivity labels and Data Loss Prevention policies for four weeks starting January 2026, allowing unauthorized access to confidential emails and highlighting security vulnerabilities in AI inference pipelines. - VentureBeat

Expecto Plagiarism! Harry Potter and the Prisoner of Fair Use: Some guy uploaded Harry Potters to Kaggle for an AI project, then someone at Microsoft pointed to them as an example of how you could train your AI, then, things went a little sideways - Ars Technica

Git Blame AI: Microsoft removed an AI-generated flowchart from its GitHub documentation after Vincent Driessen identified it as a plagiarized version of his 2010 branching model containing nonsensical errors and lacking proper attribution. - PC Gamer

Roadmap to irrelevance? Notion CEO: “If your product cannot be used by agents, I don’t think the future is very promising for you.” - Sources

Apple researchers developed Ferret-UI Lite, a 3-billion parameter model that uses real-time cropping to navigate user interfaces on-device, matching the performance of models 24 times larger on benchmarks like AndroidWorld. - 9to5Mac

Andrej Karpathy tweeted that "Claws" are a new layer for LLM agents adding orchestration, scheduling, context, tool calls and persistence, citing OpenClaw and NanoClaw (~4,000 lines) and noting Mac Minis are selling strongly. - Simon Willison’s Weblog

Karpathy: First there was chat, then there was code, now there is claw. Nanoclaw(s) maybe more interesting that OpenClaw. - X (formerly Twitter)

Sandbox or be sandbagged: Security experts recommend using Docker sandboxing, closed networks, and human-in-the-loop approvals to protect personal AI agents from data exfiltration risks caused by the combination of private data, untrusted content, and external communication. - Ihor Katkov

Comparing SVG animations made with Gemini Pro 3.0 and 3.1 - Reddit

Tech leaders warn public and corporate AI adoption lags despite high valuations; NBER finds 80% of firms report no productivity impact, Pew (2025) says 61% want more control, and Gallup Q4 2025 reports 38% workplace integration. - The New York Times

Behind every robotaxi is a Filipino on Zoom? Newly released government documents detail the remote assistance programs used by Tesla and Waymo, clarifying how human operators support autonomous vehicle safety as robotaxi services expand. - WIRED

Résumépocalypse: Recruiters report jobseekers' heavy use of AI resume generators and auto-apply tools produced many near-identical applications; Oceans received 300 similar video responses, prompting some recruiters to deprioritize AI-generated submissions while others use AI match scores (75%+) to screen. - The Washington Post

TurboTax to TurboGuess: 46% of Americans say they trust AI for tax advice and 21% plan to use it to prepare returns, but experts warn generative models hallucinate, may use outdated training cutoffs, and recommend certified tools or professionals to avoid errors. - Quartz

Hilary Allen, on Masters in Business, discussed how artificial intelligence could freeze societal progress. - Bloomberg

Child’s Play: Tech’s new generation and the end of thinking - Harper's Magazine

Royal Caribbean uses AI to automate 90% of daily pricing and reduce food waste by 50%, while reporting $17.5 billion in annual earnings and positioning itself against land-based destinations like Las Vegas.vegas/ - Fortune

Goldman Sachs and S&P Dow Jones Indices launched the S&P ex-AI index (SPXXAI) to track S&P 500 stocks unrelated to artificial intelligence, which represent 55% of the benchmark's weight and gained 32% over the last three years. - Slashdot Linux

Fake faces, realer than you: AI-generated deepfakes of Croydon spaces have gone viral on TikTok and Instagram, been reshared by copycats amassing millions of views and monetized, prompting uneven platform enforcement and fueling racist backlash despite YouGov finding 81% say their local area is safe. - BBC

Researchers from UNSW Sydney and ANU found that AI-generated faces are perceived as more realistic than human photographs, noting that extreme symmetry and average features now challenge the detection capabilities of both average people and "super recognizers." - TechSpot

Alright alrighit AI-right: Matthew McConaughey told college students at a CNN and Variety town hall that AI likenesses are inevitable and urged performers to trademark and license their likenesses, saying he has eight trademarks and invested in Eleven Labs for voice localization. - Business Insider

AI songs make up 40% of music uploaded to streaming services. - YouTube

We need BlueSky but for music, cause this is some bullshit. If Spotify can't make a proper model to label AI, build something where everyone can make labelers, discovery, DJs etc.

James Bond x Seedance 2.0 - Reddit

This song but for AI and linkedin, and not Puerto Ricans writing letters home in the 70s about how great life in NYC is - YouTube

Follow the latest AI headlines via druce.ai on Bluesky