Meet Marc Toberoff: Copyright litigator refiled Elon Musk's 15-count federal suit against Sam Altman and OpenAI in August, including fraud and RICO claims; Judge Yvonne Gonzalez Rogers denied summary judgment and opening arguments are scheduled for late April in Oakland. - Vanity Fair

OpenAI's scramble for compute - The Information

At the India AI Impact summit, OpenAI CEO Sam Altman said claims ChatGPT uses about 17 gallons of water per query are "completely untrue" and noted global data centers consume ~415 TWh, with IEA projecting ~945 TWh by 2030. - TechSpot

From spicy coworker to HR narc: Users report newer ChatGPT prioritizes safety and agreeableness—redirecting self-harm prompts to hotlines, sometimes denying earlier statements—and reducing candid feedback, prompting frustration, use of multiple AIs as workarounds, and calls for configurable tone settings. - Reddit

Anthropic launched Claude in PowerPoint as a Microsoft AppSource add-in (research preview) that embeds Claude into active PowerPoint to generate native editable slides, support three workflows, read slide masters, and warn of beta security and compliance limitations. - AI Tools Club

U.S. Defense Secretary Pete Hegseth summoned Anthropic CEO Dario Amodei to the Pentagon on Tuesday for talks over military use of the Claude AI amid Pentagon pressure to host major models on classified networks and dispute over Anthropic's usage restrictions. - Yahoo

After deadly Maduro raid, AI power struggle erupts at the Pentagon. - The Washington Post

Dozens of state bills this legislative season are advancing bipartisan AI rules, including Florida and New York chatbot disclosure and child protections, limits on using names/images without consent, and data-center energy or moratorium measures amid industry and federal pushback. - NPR

Eduardo Porter: If labor is obsolete, who decides who gets to eat? UN Secretary-General António Guterres warned that AI-driven declines in labor income could erase wage tax revenue and concentrate power, while analysts note the US tax take is under 26% of GDP and capital taxation about 2%, urging early government action. - the Guardian

AI bubble, productivity donut: Goldman Sachs and JPMorgan economists calculated that massive AI investment contributed "basically zero" to U.S. economic growth in 2025, contradicting claims it drove as much as 92% of growth and tempering near-term GDP expectations. - The Washington Post

Basel 3 meets Gemini 3.1: The European Central Bank has increased scrutiny of European banks' exposure to the artificial intelligence industry. - Bloomberg

When technical debt and financial debt meet-cute: Software companies are delaying debt deals as higher borrowing costs and lender scrutiny meet AI disruption risk, with tech loans 17% ($260B) of leveraged loans, the software index down 20% year-to-date and 50% of software loans rated B- or lower. - Yahoo

Meta reported strong revenue but has borrowed billions to fund artificial intelligence development. - The Wall Street Journal

Hey you, keep your claws off of our cloud: Google banned OpenClaw Antigravity OAuath access - Implicator.ai

iSee dead batteries: Tim Cook's increased mentions of Visual Intelligence suggest it will be a major feature of Apple's upcoming wearables. - Bloomberg

Bloomberg's Mark Gurman reports Apple will embed its Visual Intelligence across upcoming AI wearables — smart glasses targeting a 2027 launch (production possibly Dec 2026), an AI pin possibly in 2027, and AirPods as soon as this year. - MacRumors

Samsung is integrating Perplexity into Galaxy AI on Galaxy S26, accessible by saying 'hey, Plex,' with OS-level access to Samsung Notes, Clock, Gallery, Reminder and Calendar, and select third-party apps, and will discuss details at Unpacked event on the 25th. - The Verge

Google for Education, with ISTE and ASCD, will offer free Gemini-focused AI literacy training to all 6 million U.S. K–12 and higher-education faculty—covering 74 million students—using microcredentials, bite-sized modules, and tools like Gemini, NotebookLM, and Guided Learning. - Google

Washington Park High School in Newark launched an A.I. literacy course for seniors to teach them to actively use and critique AI tools and prepare for an AI-shaped workforce. - The New York Times

Retail therapy, now with AI: Retailers are adopting artificial intelligence to transform shopping and customer experiences. - Slate Magazine

Interpersonal and strategic leadership skills cannot be replicated by AI. - Fast Company

Real-estate brokers are expressing anxiety about how artificial intelligence will affect their industry. - The Wall Street Journal

If you perform no useful service, can AI disrupt you?

New AI tools promise to automate inbox management and could reduce the productivity burden of email, but their effectiveness remains uncertain. - Financial Times

Teenage mutant ninja founders are launching AI startups across consumer and health, with Cal AI reaching eight million downloads and ~$30M annual revenue run rate and Synthetic Sciences raising $1.5M (including $500K from Y Combinator), prompting VC programs to adapt. - New York Post

A report finds 82% of organizations use AI agents, 80% say agents performed unauthorized actions, 92% of tech leaders call governance crucial, only 44% have policies, and 98% plan to expand deployments. - TechRadar

Multi-step AI agents suffer compounding error—90% per-step accuracy yields ~12% end-to-end success over 20 steps—so production favors constrained or hybrid workflows, external validation, and human checkpoints amid higher operational costs (agent runs ~$230–$360 per attempt). - Hacker Noon

Kinda works? The article defines five agent-focused evaluation metrics—Task Completion Rate, Tool Selection Accuracy, Autonomy Score (Human Intervention Rate), Recovery Rate, and Cost per Successful Task—to assess AI agents' task success, tool use, autonomy, error recovery and cost efficiency. - Machine Learning Mastery

AI-driven querying multiplies DirectQuery warehouse queries into dozens or hundreds, causing performance slowdowns and higher cloud costs, and vendors recommend using Import mode for cached decision cadences or Direct Lake/governed ingestion pipelines (e.g., Microsoft Fabric) to scale. - Hacker Noon

Infosys chairman Nandan Nilekani said AI makes retaining legacy systems untenable, noting many large firms spend 60–80% of IT budgets on maintenance, and warned AI will create new 'AI legacy systems' that will require cleanup and governance. - The Register

AI still can’t read PDFs: Public document releases—House Oversight Committee's 20,000‑page Jeffrey Epstein files and the DOJ's more than three million PDFs—used poor OCR that left documents effectively unsearchable, highlighting persistent AI failures to extract structured data from PDFs. - The Verge

Seagate and Western Digital said their nearline HDD capacity is fully allocated through calendar 2026 as hyperscalers buy high-capacity drives, leaving little open production for discretionary buyers and prompting shortages, price increases and shifted enterprise storage forecasts. - The Register

The new cutting the brake line: Georgia Tech researchers investigated VillainNet, a hack which would enable a bad actor to poison "super network" architectures for self-driving cars with a dormant backdoor that succeeds 99% when triggered, and evade current security tools. - Digital Trends

An experimental hybrid trial led by Cyborg Dynamics Engineering and Griffith University using one physical UGV plus up to four virtual teammates showed a 99.67% success rate navigating obstacles and extinguishing two simulated fires, trained via multi-agent reinforcement learning. - Interesting Engineering

Team Mirai, a party founded by software engineers, won 11 of 465 parliamentary seats in Japan by promising self-driving buses and high-tech jobs. - The New York Times

President Donald Trump shared an AI-generated video on Truth Social, posted by Kari Lake, showing him confronting Canada's hockey team captioned 'Trump, THE ENFORCER' as U.S. tourists sheltered in Mexico amid violence after a drug lord's killing. - The Daily Beast

AI necromancy: Venture capitalist John Arrow launched "AI Coffee with Scott Adams" after Scott Adams's death, using a deepfake model trained on his transcripts and writings, against Adams's family wishes, but says Adams himself gave the green light. - Business Insider

11th commandment: Thou shalt not copy-paste? Pope Leo XIV told priests of the Diocese of Rome to rely on their own judgment rather than artificial intelligence when preparing homilies, and urged sustained prayer, ongoing study, fraternity and care for lonely elderly priests. - EWTN News

Is this AI? - Reddit

Follow the latest AI headlines via druce.ai on Bluesky