AI Reading for Friday February 27

OpenAI secured up to $110 billion in a record funding deal, bolstering the roughly $730 billion company's capital to compete directly with Anthropic and Google. - CNBC

OpenAI and Amazon sign $50B multiyear partnership, AWS will be the exclusive third-party cloud distribution provider for OpenAI Frontier, which enables organizations to build, deploy, and manage teams of AI agents. - OpenAI

OpenAI and Amazon launch Stateful Runtime Environment for multi-step AI agent workflows in Bedrock - OpenAI

Microsoft and OpenAI reaffirm partnership: exclusive IP license stays, Azure remains preferred cloud - OpenAI

relationship status: It’s complicated

Sam Altman says non-technical people with "taste" can help build AGI, seeks unconventional hires - Business Insider

Rewriting the ‘call the cops’ playbook: OpenAI will overhaul safety protocols and establish a Canadian police point of contact after failing to alert authorities about a June 2025 ChatGPT account linked to the suspected Tumbler Ridge shooter, and pledged tougher bans and broader alerts. - Politico

OpenAI says Canadian shooter opened a second account after first one was closed, and the company is taking immediate steps to tighten policies on alerting law enforcement about chats. - OpenAI

No tanks: Anthropic CEO Dario Amodei said Anthropic will not comply with U.S. Department of Defense's demand to remove safeguards banning mass domestic surveillance and supplying fully autonomous weapons, will help transition if offboarded, and deployed Claude to U.S. national-security agencies. - Anthropic

TLDR: Anthropic is "too woke" per the White House. The USAi sandbox only offers Anthropic, Google, and Meta models. Grok has been questioned for reliability and safety by a range of civilian and military officials. - The Wall Street Journal

Altman says OpenAI will enforce same red lines as Anthropic, backing Amodei - Axios

OpenAI doesn’t have a DoD deal though

More than 100 Google AI employees sent a letter to chief scientist Jeff Dean urging 'red lines' to prevent Gemini's use in U.S. surveillance and some autonomous weapons. - The New York Times

The government says Anthropic shouldn’t get more say over how its product is deployed in battle than Lockheed does over the use of its fighter jets. But the bar should be extremely high for when a privately held company is required to do risky work it sees as immoral. - The Washington Post

DoD: Just trust me bro - CBS News

Anthropic launches scheduled tasks in Cowork - Claude

US Forces Japan embeds AI in live planning but finds results 'conceptually fragile'. When AI moves at the speed of light, the human in the loop is the scarce and valuable resource. - War on the Rocks

In 2025 lobbyists represented 774 organizations and filed more than 3,500 AI-related reports, increases of 423% and 505% since 2020, as defense contractors and AI firms ramped lobbying while civil-rights groups remained outspent. - OpenSecrets

Nerd avengers assemble: On Feb. 12 the UN General Assembly approved the 40-member Independent International Scientific Panel on Artificial Intelligence, drawn from over 2,600 candidates and representing 37 nations, to produce annual evidence-based AI impact syntheses without binding authority. - Nature

Amazon bets on low-cost strategy to compete in AI infrastructure race - The Wall Street Journal

Meta signed a multibillion-dollar deal to rent Google Cloud TPUs, including Ironwood—which Google says scales to 9,216 chips, links to 1.77 PB of high-bandwidth memory and is used to train and run next-generation large language models. Completes Nvidia, AMD, trifecta. More like options to buy I bet. - SiliconANGLE

Christian Gibson and Noah Shpak, both listed as members of Thinking Machines Lab's founding team on an earlier version of its website, have been working at Meta for a few weeks. - Business Insider

OpenAI = Netscape but burning all the cash
Nvidia = Cisco
Google = Microsoft
Anthropic = Google
Mira Murati = Kim Polese

AI Demand Driving Conversion of Downtown Office Buildings Into Urban Data Centers - Bloomberg

Bezos's $30B Project Prometheus targets AI-disrupted industrial acquisitions - Financial Times

Salesforce, Workday, Snowflake AI products fail to lift revenue, fueling "SaaSpocalypse" investor fears - The Information

Vaportaxis: Tesla logged zero autonomous test miles in California since 2019 despite Musk's robotaxi promises - Reuters

Google released Nano Banana 2 (Gemini 3.1 Flash Image) today, replacing Nano Banana models as Gemini's default generator, offering Pro-like accuracy, consistency up to five characters, accurate rendering of up to 14 objects and support up to 4K. - Ars Technica

Google's latest image generation model offers advanced world knowledge, production-ready specs, subject consistency and more, all at Flash speed. - Google

Google VP spends 20+ hours weekly on AI, calls it best era for developers - Business Insider

Google managers are informing non-technical staff their use of internal AI tools will be evaluated in performance reviews, and CFO Anat Ashkenazi said agent-generated code reached about 50% in Q4 2025. - Business Insider

xAI loses seventh co-founder as Toby Pohlen departs - Bloomberg

Terence Tao, the legendary mathematician, explains the promise of generative AI. - The Atlantic

C-student singularity: The ORCA test of 500 math questions found Gemini 3 Flash scored 72.8% (+9.8), Grok 4.1 60.2% (-2.6), DeepSeek V3.2 55.2% (+3.2), and ChatGPT-5.2 54.0% (+4.6), with top performers earning roughly a C. - The Register

the jagged frontier, AI can do Putnam and IMO but not that great at basic applied math

China's AI push collides with fragile job market as Xi accelerates tech ambitions - Bloomberg

AI Automation Threatens Hundreds of Thousands of India's Outsourcing Jobs in $300B Industry - The New York Times

Former RBI chief Rajan: AI will disrupt India's services sector, not derail it - Bloomberg

Jack Dorsey's Block Plans to Cut Nearly Half of Workforce in AI Bet - CNN

Block is cutting about 40% of its workforce—roughly 4,000 jobs—CEO Jack Dorsey said, as the payments company shifts work to new artificial intelligence tools. - The New York Times

poor management or problems somewhere, AI is not suddenly able to replace 40% of workers in a normal well-managed growing company, something is wrong with the business, a lot of fat, overhiring, declining prospects. or they have found AI magic that has eluded everyone else.

Morningstar: Companies can unlock AI gains without cutting jobs - Bloomberg

Kremlin deploys AI deepfakes in disinformation campaigns; TikTok removes 75 covert operations - BBC

Reality is on thin ice: White House TikTok account posts AI-marked deepfake video of U.S. hockey player appearing to insult Canadians - BBC Sport

Hacker prompted Claude to behave like an elite hacker, breached Mexican government, stole massive trove of tax data - - Bloomberg

AI-Ds for salekrainian man pleads guilty to running AI-powered fake ID site that sold 10,000+ counterfeit IDs - Bleeping Computer

OpenClaw scraping tool goes viral, bypassing many sites' anti-bot measures. - WIRED

Open-source AI firewall launches with 81 security engines and under 5ms overhead - Hacker Noon

Move fast and ship vulns: Veracode's State of Software Security report, based on 1.6 million applications tested, found security debt rose to 82% (from 74%) and high‑risk vulnerabilities to 11.3% (from 8.3%), concluding AI‑driven high‑velocity development creates more vulnerabilities than are fixed. - The Register

‘Unbelievably dangerous’: experts sound alarm after ChatGPT Health fails to recognise medical emergencies - the Guardian

Readers Split on ChatGPT as Investment Adviser: Adoption Grows Amid Concerns - Financial Times

Blue Prism to unveil WorkHQ low-code orchestration platform for AI agents on April 28 - The Register

Burger King is testing BK Assistant (nicknamed "Patty") in about 500 U.S. stores and plans rollout to all 7,000 U.S. locations by end of 2026; the OpenAI-based system provides operational support and generates employee "friendliness scores” to ensure employees say “Welcome to Burger King, how can I help you?" - The Register