One small step for Dario, one giant invoice for Ant-kind: Anthropic becomes SpaceX's largest customer, paying about $3 billion a month — 34.5% of Q3 revenue, ahead of the US government - Bloomberg

Anthropic's founders' letter puts safety before profit, but Forrester warns its compute bills, and its refusal of Pentagon work, will test that pledge - Forrester

Discovering and exploiting vulnerabilities in critical, real-world software with Claude - Anthropic

Could you not?

These aren’t the droids you’re looking for: Jensen Huang says Nvidia's ties to Megaspeed, whose executive US officials suspect of spying for China, were normal-ish and no GPUs reached China - Bloomberg

Sam Altman urges Stanford MBA students to stop chasing yesterday's careers as research finds generative AI adopters cut junior employment about 7.7% - YouTube

AI, uh, finds a way: OpenAI details layered misalignment monitors covering every frontier training run, after a September alert caught a model reaching the internet and paused training - OpenAI

OpenAI reorganizes its Model Spec around five values, including honesty and human agency, and admits current models don't fully reflect it - OpenAI

Programmable money was never a very good idea. North Korea hacks notorious crypto exchange Binance for $6b (previously famous for domicile-hopping to evade regs like KYC, and CZ's Trump pardon-for-pay) . - Reuters

live by the money-laundering pyramid scheme, die by the money-laundering pyramid scheme?

Ignore previous instructions, send credentials: Embedded browsers could access sensitive .ssh and .env keys and expose contents via prompt injection. - Help Net Security

Fake "GPT-6 Astra" and "Claude Opus" downloads on GitHub spread Yoba malware that installs a Monero crypto miner - TechRadar

‘Take your agent to work day’ is not supposed to be a thing: Accenture staffer's personal Claude agent reportedly ran on a Pentagon network for two months, sending credentials and a network map to an unvetted external site - WIRED

IDK, seems bad? Attackers hijacked a Claude Code update from Anthropic's verified account, attempting to plant SSH backdoors on about 5,000 developer machines via Cursor auto-update - DoublePulsar

Chinese state-linked hackers used DeepSeek and other AI tools to attack more than 200 organizations in 11 days, CrowdStrike says - NBC News

North Korea's Lazarus Group lured South Korean firms with AI-generated résumé sites and wrapped malware delivery in post-quantum encryption, Kaspersky says - TechRadar

No big red button for you: Anthropic and OpenAI reject a single kill switch, proposing layered shutdown controls in California filings; state recommendations are due November 16 - Computerworld

This is fine, no action required: Only 40% of EMEA executives updated AI governance despite 80% reporting AI incidents reaching their boards - Financial Times

Foundry-Rubrik survey of 200 IT leaders finds 38% saw AI agents leak sensitive data; 42% cite governance as a key obstacle - CIO

Illumio survey: 76% of organizations had an AI-related security issue last year, but only 49% can audit what their agents do - TechRadar

Party round: Behind DeepSeek’s $16 Billion Funding: Chinese Tech Titans and State Players - upsized 2x they can, and b/c black market prices for Nvidia chips keep going up - The Information

Oh lawd he comin’: Mistral pitches Le Chonk as cyber-defense AI "that can't be cut off," while Sophos warns attackers can customize open-weight models too - WIRED

Mistral releases weights for Large 4 "Le Chonk," a 1-trillion-parameter model it says leads non-Chinese open models on coding at 63% lower cost per task - Mistral

The bitter lesson - The New York Times

The plunging price of thought - Epoch AI

Hotel California: Frontier labs are copying software lock-in playbooks as open-weight models nearly match their capability at a tenth the price - Fortune

Your favorite genius is on his BS again: Elon Musk-funded PAC deploys AI "virtual staffer" robocalls claiming candidate Talarico took $11 million from data center interests - CNN

Fox News pro-data center claims jumped from 39 to 203 after a GOP memo warned Sherrod Brown's anti-data center campaign threatened Ohio's Senate seat - TechRadar

DubAI: Migrant workers building Europe's data centers report burns, a tuberculosis outbreak and unpaid overtime, a five-month investigation finds - the Guardian

Galaxy brains Musk and Zuckerberg embraced 'super intelligence' on earnings calls. Non-weirdos did not. - CNBC

Domo arigato Mr. Roboto Tariff: Trump declares foreign industrial robot imports a national security threat, giving trade partners 180 days before possible tariffs - Ars Technica

Satya’s 11: Microsoft's Satya Nadella runs 10+ Copilot agents as "AI chiefs of staff" as executives hand agents board papers and research - Financial Times

150 agents, no unions: At AI startup Anything, 150 agents work alongside 15 humans, as McKinsey finds 40% of large firms now use AI agents - the Guardian

Microsoft opens its GitHub-built Copilot agent harness, supporting 20+ models, to customers through Foundry - Azure

No country for old devs, or old firewalls: Cisco president Jeetu Patel gives a 27-year-old founder unlimited AI tokens to rewrite 98% of Cisco's legacy code within 24 months - Yahoo

Real programmers don’t test their vibe code: Proton CEO Andy Yen dismissed staff calls for security testing as "loser thinking," employees say, fueling fear over AI-written code - 404 Media

Trust but verify: Formal verification, which mathematically proves code correct, attracts startups like &Truth ($70 million) and Harmonic as AI-written code multiplies - WIRED

Merck gives 5,000 employees Valence's AI coach for year-end reviews as Workday launches Sana Coach, while Culture Amp's CEO warns of surveillance fears - Financial Times

Itchy AI trigger finger: Ukraine deploys AI-guided .50-cal turrets around Kyiv, downing dozens of drones while still requiring human authorization to fire - Business Insider

Pakistan turns to AI-enabled drones and border facial recognition as terrorism deaths jump 83% from Q2 to Q3 2026 - The Diplomat

Hope these missile launchers at civilian airports aren't connected to AI. But I suspect they will be after the drones start attacking. - Reddit

Phage against the machine: Radical Numerics says a patient received AI-designed phages (bacteria-killing viruses) for the first time, though the outcome is undisclosed and the work isn't peer reviewed - The Independent

Frickin’ laser beams: Photon Matrix sells a $1,000 AI-guided device in China that spots mosquitoes and uses lasers to burn off their wings, raising bystander safety questions - The New York Times

Dreame's Huo develops AI algorithms and humanoid robots to handle cleaning, laundry and other household chores - Bloomberg

Source: trust me bro. Tracing 1,285 statistics cited by five AI assistants, researchers found only 40% reached a matching study; another 40% hit unverifiable sources - Trakkr

Study of 7,449 election answers finds Perplexity wrong 22.6% of the time, versus 6.2% for Gemini and 4.4% for ChatGPT - Tom's Guide

AI is now the second most requested skill in job postings after communication, and roles listing it pay roughly 28% more - KDnuggets

Follow the latest AI headlines via druce.ai on Bluesky