OpenAI's contract with the U.S. Department of Defense permits cloud-only deployment of AI in classified environments, retains its safety stack and cleared personnel, and enforces three red lines: no mass domestic surveillance, no autonomous-weapon direction, no high-stakes automated decisions. - OpenAI

Mr. Altman agreed to the use of OpenAI’s technology for all lawful purposes, but also negotiated the right to put safeguards into OpenAI’s technologies that would prevent its systems from being used in ways that it did not want them to be. - The New York Times

Bringing the smarm to a Twitter AMA - X (formerly Twitter)

Altman explains - - X (formerly Twitter)

basically, we gave them everything they wanted with a fig leaf that we might have red lines in the future, and of course they can designate us a supply chain risk and invoke DPA if we don’t.

unbelievably smarmy to take Feds' money with loopholes they could drive an aircraft carrier through, and then give lip service to how he is trying to support Anthropic and help de-escalate

I can't imagine why people accuse Slippery Sam of talking out of both sides of his mouth?

Chocolate covered capitalism: Hershey's corporate structure is the closest American analogue to OpenAI's unusual charitable structure, illustrating how that setup can shape governance and risk. - Financial Times

Pentagon used Claude in Iran strikes hours after Trump banned Anthropic tools - The Wall Street Journal

Trump Orders Federal Agencies to Stop Using Anthropic AI; Altman Warns of 'Scary Precedent' - International Business Times

After OpenAI agreed to let the U.S. Department of Defense deploy its models on classified networks, some users defected to Anthropic and its Claude chatbot reached No. 1 on the App Store. - Business Insider

Anthropic's Claude adds a memory-import workflow that transfers preferences, projects and conversational context from other AI providers via a two-step copy-and-paste into Claude's memory settings, available on all paid plans, to lower switching costs and speed onboarding. - Claude

Anthropic launched wealth-management plug-ins for its Claude AI, enabling advisory firms to build private, customizable programs with firm data and compliance controls and announced integrations with LPL Financial and Orion; Cerulli reports 38% of affluent investors are comfortable with AI. - The Daily Upside

Cobye-bye: Anthropic's Claude Code automates COBOL analysis—mapping dependencies across thousands of lines, generating tests, translating code and creating API wrappers—reducing modernization time and costs and prompting a drop in IBM's stock. - TechRadar

Panic at the software disco: In early 2026, generative AI expanded into agentic executive-assistant systems, triggering a market sell-off across software, legal, insurance and cybersecurity, escalating safety clashes (Anthropic blacklisted) and a $125 million super PAC targeting NY Assemblyman Alex Bores. - CNBC

Bill Gurley warned on the On with Kara Swisher podcast that workers 'sitting idly' without purpose are most at risk of being replaced by AI as Meta, Microsoft, Amazon and Alphabet invest hundreds of billions in AI infrastructure. - Business Insider

Prime cuts: Amazon layoffs leave surviving staff with guilt and heavier workloads as AI integration accelerates - Financial Times

A big AI worry? It will destroy old jobs faster than it creates new ones. The only thing lacking is proof. - Morningstar

Might happen, might not happen. But proof is a thing in math, not in social science. There are only models, sometimes they work, and sometimes they don't.

The models, depending how you parameterize them, can shrink the labor share massively or just a little. but hard to say AI increases it.

Citrini Research's viral AI report argues that doomer narratives reflect college-educated professionals' anxiety, which could drive regulatory pressure, slow AI adoption, and increase demand for targeted reskilling. - The Washington Post

Lloyd Blankfein discussed AI, private credit, and politics. - Bloomberg

Mike Green, chief market strategist at Simplify, warned global passive AUM rose 400% from 2012–2023 (LSEG), inflating US equity valuations ~15% annually and risking a 1929‑style crash if flows reverse, with private‑credit (software ~40% of PE loans) posing contagion risk. - Business Insider

On Feb. 26, 2026 former hedge fund manager Enrique Abeyta said in a presentation that AI is shifting from experiments into core industry infrastructure, altering institutional decision-making, work organization and productivity across sectors and prompting market and strategic changes. - GlobeNewswire News Room

NCMEC received over one million reports tied to generative-AI CSAM between January and September 2025, while Homeland Security Investigations reported over a 600% increase in AI-linked child-exploitation reports in the first half of 2025. - NBC News

Truffle Security found 2,863 public Google Cloud API keys that can authenticate to Google’s Gemini endpoints and expose uploaded files, cached data, and enable billing abuse, while Google said it has implemented measures to detect and block leaked keys. - The Hacker News

AI 'faith healer' deepfakes flood Meta Reels, reaching millions of followers - Futurism

that is one horrible used case and one cursed illustration

Face-off where you always lose: Royal Society testing found ordinary people correctly identify AI-generated faces only 31% of the time (super-recognizers 54%), while realistic generators have spurred commercial detection services amid documented 2024 harms including a $25 million Hong Kong fraud. - New York Post

Stanford and Princeton researchers found Chinese AI chatbots are more likely than Western counterparts to avoid politically sensitive questions or provide inaccurate answers, indicating built-in self-censorship that may erode user trust and limit international adoption. - WIRED

A self-service AI mental health field test used templated PHQ‑9/GAD‑7 prompts across major LLMs (ChatGPT, Claude, Gemini, Grok), producing plausible screening questions and referral suggestions but highlighting false negatives/positives, privacy and hallucination risks and urging safeguards. - Forbes

Siri-otypes: Many consumer AI assistants default to female voices, driven by biased training data and historical feminization of helper roles, a 2024 study warned it reinforces stereotypes, while women hold 22–26% of AI roles and under 15% of senior AI leadership. - TechRadar

Condé Nast CEO Roger Lynch warned AI is a "death blow" to Google search and said the publisher increased revenue last year driven by subscription growth. - Financial Times

The 184-year-old Cleveland Plain Dealer began publishing AI-drafted local articles under bylines like "Advance Local Express Desk" at a top editor's push, boosting web traffic while prompting staff concern about quality and reputational risks. - The Washington Post

Tech firms are enforcing rules on artificial intelligence use via company policies and oversight. - The Wall Street Journal

Apple to replace Core ML with 'Core AI' framework for on-device generative AI at WWDC 2026 - Apple Insider

Google's NotebookLM adds Gemini-powered research assistant and Deep Research mode - Tom's Guide

Personal proect: Apropos NotebookLM, I made a pipeline for writing reports, updates one I did earlier to use Claude Code. Write a DAG in YAML describing a pipeline: 1) run some scripts to gather data 2) outline sections and run a critic-optimizer loop for each section 3) assemble everything via templates - GitHub

Example report on MongoDB stock.

Example YAML DAG

super POC but combining NotebookLM to read sources and ideate how to structure, and then a structured process to research and write is powerful for e.g. literature review. Could use Zotero or e.g. Letta as memory.

China's smartphone maker Honor demonstrated a humanoid and a robot phone prototype in an AI pivot. - Bloomberg

HONOR unveiled the Robot Phone at MWC Barcelona, a smartphone with an integrated motorized AI-driven camera gimbal that provides automated subject tracking and intelligent framing, billed as a "personal cinematographer". - Mashable

Follow the latest AI headlines via druce.ai on Bluesky